JP5390458B2 - Moving picture predictive coding apparatus, moving picture predictive coding method, moving picture predictive coding program, moving picture predictive decoding apparatus, moving picture predictive decoding method, and moving picture predictive decoding program - Google Patents

Description

  The present invention relates to a moving picture predictive coding apparatus, method, and program, and a moving picture predictive decoding apparatus, method, and program.

  In order to efficiently transmit and store moving image data, a compression encoding technique is used. In the case of moving images, MPEG1-4 and H.264 are used. 261-H. H.264 is widely used.

  In these encoding methods, an image to be encoded is divided into a plurality of blocks, and encoding / decoding processing is performed on each block. In order to increase the encoding efficiency, the following predictive encoding method is used. In predictive coding between screens, the displacement of a signal is searched with reference to a previously reproduced image signal in a screen different from the target block, a predicted signal is generated by compensating for the displacement, and the predicted signal is converted to the target block. The difference signal obtained by subtracting from the signal is encoded. Recent techniques use motion search with increased accuracy of signal displacement. An already reproduced image that is referred to for motion search / compensation is referred to as a reference image.

  In addition, in the inter-screen prediction, not only the past image displayed before the target image in the display time order but also a future image displayed after the target image may be referred to together (however, Future images need to be encoded and reproduced in advance before the target image). By averaging both the prediction signal obtained from the past image and the prediction signal obtained from the future image, it is effective for the prediction of the signal of the hidden and newly appearing object, and at the same time both predictions This has the effect of reducing noise contained in the signal.

  Further, H.C. In the H.264 inter-frame prediction encoding, the prediction signal for the target block refers to a plurality of reference images encoded and reproduced in the past, and an image signal with the least error is selected as an optimal prediction signal while performing motion search. . Then, a difference between the pixel signal of the target block and the optimal prediction signal is obtained, and the difference is subjected to discrete cosine transform and quantized, and then entropy-coded. At the same time, information relating to which reference image the optimum prediction signal for the target block is acquired from (reference index) and information relating to which region in the reference image the optimum prediction signal is obtained (motion vector) are also encoded. To do. H. In H.264, 4 to 5 reproduced images are stored in the frame memory as reference images. In this specification, the frame memory includes a so-called decoded picture buffer.

として表現される。ただし、画像はN個のブロックに分割されているとして、B_iは画像のi番目のブロックに含まれる画素を意味し、(mvx,mvy)はi番目のブロックの動きベクトルを意味する。 Here H. As a motion compensation method employed in H.264 and MPEG4, a block matching method capable of fractional motion compensation such as ½ pixel accuracy and ¼ pixel accuracy is employed. When this method is expressed by a mathematical expression, it is expressed as follows. The pixel value (luminance value or color difference value) in the coordinate (x, y) of the predicted image P of the image frame (target frame) to be encoded is P (x, y), and the coordinate (x, y) of the reference image R The pixel value at is defined as R (x, y). Here, assuming that x and y are integers, it is assumed that a pixel exists at a point where the coordinate value is an integer in P and R. At this time, the relationship between P and R is

Is expressed as However, assuming that the image is divided into N blocks, B _i means a pixel included in the i-th block of the image, and (mvx, mvy) means a motion vector of the i-th block.

と表現される。ただし「//」は、除算の結果を近隣の整数に丸め込むことを意味する。 If the motion vector (mvx, mvy) is not an integer, a sample value that does not actually exist in the reference image is calculated. Here, H. For the H.264 color difference signal, interpolation using the color difference signals of the surrounding four pixels is performed. When this interpolation method is described by a mathematical expression, d is a positive integer, 0 ≦ p ≦ d, 0 ≦ q ≦ d, and the interpolation sample value R (x + p / d, y + q / d) is

It is expressed. However, “//” means that the result of division is rounded to a neighboring integer.

  At this time, if the interpolated sample value R (x + p / d, y + q / d) is not an integer, an error occurs due to the rounding process of division. In general, in a situation where sufficient bits for encoding the inter-frame prediction error cannot be secured, such as when encoding at a low bit rate, the quantization step size of the discrete cosine transform is set large. For this reason, there is a case where an error generated by motion compensation cannot be corrected by error coding. Further, even if a residual signal is sent, quantization distortion is included, so it is difficult to completely eliminate the error. Here, since a reproduced image including these errors is used as a reference image for motion compensation, when the inter-frame coding continues without performing the intra-frame coding, the above error accumulates, and temporally Objective performance such as PSNR of the color difference signal is degraded, and a phenomenon occurs in which the color of the image is subjectively biased to red or green. In particular, since the human eye is sensitive to red, when the video turns red due to the above error accumulation, the subjective quality is deteriorated.

  In Patent Document 1 below, a method of rounding up a value obtained by adding 0.5 to the result of the interpolation method for calculating the sample position indicated by the fractional precision motion vector when performing motion compensation prediction, and the result of the interpolation method described above. There is a description of a method of switching the method of truncating after adding 0.5 for each image constituting the video.

  Further, in Patent Document 1, as a method of suppressing reddening of a reproduced video, an “implicit method” in which a rounding method is alternately changed over the entire screen based on the temporal generation order of frames, and a reference image is predicted and generated. There is an “explicit method” that stores the rounding method used and uses a rounding method different from the stored rounding method as the rounding method for the entire reference image when predicting the target image. Have been described.

JP 2008-160877 A

  H. In the H.264 inter-screen predictive encoding, a plurality of reference images that have been encoded and reproduced in the past are referred to in order to generate a prediction signal for each block of the target frame. For this reason, in the implicit method of Patent Document 1, the rounding method may not be switched appropriately.

  This will be described with reference to FIG. In FIG. 1, when performing inter-frame prediction of the target frame f105, the reference images f101 and f103 are used when the rounding method A (for example, a method of rounding up a value added with 0.5) is used when the target frames are predicted. f102 and f104 indicate a situation in which a rounding method B (for example, a method of rounding off after adding 0.5) is used when predicted as a target frame. At this time, according to the implicit method of Patent Document 1, inter-frame prediction is performed on the target frame f105 using the rounding method A. Here, it is effective when the prediction is performed using the reference images f102 and f104 generated by the rounding method B, but the same rounding method is used when the prediction is performed using the reference image f103 generated by the rounding method A. A is used. Therefore, the same rounding method is used and the effect of suppressing the accumulation of rounding errors is lost.

  Even in the explicit method described in Patent Document 1, even when the rounding method for the entire reference image is changed, if the prediction is performed using a plurality of reference images, the randomness of the rounding method is lost. If errors are accumulated in the entire screen, the image quality of the entire screen is degraded.

  In order to solve the above-described problems, the present invention suppresses the accumulation of rounding errors, improves the quality of a reproduced image, improves the prediction efficiency of an image that uses the reproduced image as a reference image, and improves the subjective performance. Objective.

In order to achieve the above object, a video predictive coding apparatus according to an aspect of the present invention includes an input unit that inputs a plurality of images constituting a video and intra-screen prediction or motion compensation for the input images. Encoding means for generating compressed image data by encoding according to any one of the inter-screen prediction using the image, and a restoring means for restoring the reproduced image by decoding the generated compressed image data Necessary for the fractional accuracy prediction when the reproduced image is stored as a reference image used for encoding the subsequent image and the input image is encoded by the inter-screen prediction. comprising a rounded method determination means for determining the rounding method of the sample values generated by a pixel interpolation, wherein the rounding method determination means, when performing case and bidirectional prediction for performing unidirectional prediction In individually, depending on the sample position interpolated in the reference image, and determines the rounding method to be performed in the sample position. In addition, the moving image predictive coding apparatus according to one aspect of the present invention includes an input unit that inputs a plurality of images constituting a moving image, and inter-screen prediction using intra prediction or motion compensation for the input image. Encoding means for generating compressed image data by encoding any of the above, a restoring means for restoring a reproduced image by decoding the generated compressed image data, and the restored reproduced image, Image storage means for storing a subsequent image as a reference image used for encoding, and when the input image is encoded by the inter-screen prediction, generated by pixel interpolation necessary for fractional accuracy prediction Rounding method determining means for determining a rounding method of sample values to be performed, and the rounding method determining means is used for two types of prediction signals used in bi-directional prediction. Individually for the image, depending on the sample position interpolated in the reference image, and determines the rounding method to be performed in the sample position. At this time, the rounding method determination unit may determine the rounding method based on information on a rounding method used when the reference image is predicted as a target image.

In addition, the moving image predictive coding apparatus according to one aspect of the present invention includes an input unit that inputs a plurality of images constituting a moving image, and inter-screen prediction using intra prediction or motion compensation for the input image. Encoding means for generating compressed image data by encoding any of the above, a restoring means for restoring a reproduced image by decoding the generated compressed image data, and the restored reproduced image, Image storage means for storing a subsequent image as a reference image used for encoding, and when the input image is encoded by the inter-screen prediction, generated by pixel interpolation necessary for fractional accuracy prediction anda method determining means rounding to determine how rounding sample values, the rounding method determination means, separately for the case where the case and bidirectional prediction for performing unidirectional prediction, reference image How rounding said to be performed at the sample position interpolated to definitive, the reference image and determines based on the information on how the rounding is used when the predicted as the target image. In addition, the moving image predictive coding apparatus according to one aspect of the present invention includes an input unit that inputs a plurality of images constituting a moving image, and inter-screen prediction using intra prediction or motion compensation for the input image. Encoding means for generating compressed image data by encoding any of the above, a restoring means for restoring a reproduced image by decoding the generated compressed image data, and the restored reproduced image, Image storage means for storing a subsequent image as a reference image used for encoding, and when the input image is encoded by the inter-screen prediction, generated by pixel interpolation necessary for fractional accuracy prediction Rounding method determining means for determining a rounding method of sample values to be performed, and the rounding method determining means is used for two types of prediction signals used in bi-directional prediction. The rounding method to be performed individually at a sample position to be interpolated in a reference image is determined based on information on a rounding method used when the reference image is predicted as a target image. And

The rounding method determining means of the moving picture predictive encoding apparatus may determine the rounding method individually for each reference image .

  Further, the rounding method determining means may determine the rounding method for at least one of the luminance signal and the color difference signal. Further, the rounding method determining means may determine the rounding method individually for each of the luminance signal and the color difference signal.

To achieve the above object, a video predictive decoding apparatus according to an aspect of the present invention encodes a plurality of images constituting a video by either intra prediction or inter prediction using motion compensation. Input means for inputting the compressed image data including the motion vector, a restoration means for restoring the reproduced image by decoding the compressed image data, and the restored reproduced image as a subsequent image. Generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-frame prediction including fractional accuracy prediction and image storage means for storing as a reference image used for decoding comprising a rounded method determination means for determining the rounding method of the sample values, wherein the rounding method determination means, separately for the case where the case and bidirectional prediction for performing unidirectional prediction, reference image Depending on the sample position interpolated to definitive, and determines the rounding method to be performed in the sample position. The moving picture predictive decoding apparatus according to one aspect of the present invention is obtained by encoding a plurality of images constituting a moving picture by either intra prediction or inter prediction using motion compensation. , Input means for inputting compressed image data including a motion vector, restoration means for restoring a reproduced image by decoding the compressed image data, and use of the restored reproduced image for decoding subsequent images And a method of rounding a sample value generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction. Rounding method determining means for determining, and the rounding method determining means individually for reference images used for two types of prediction signals used when bi-directional prediction is performed, Depending on the sample position interpolated in irradiation image, and determines the rounding method to be performed in the sample position. At this time, the rounding method determination unit may determine the rounding method based on information on a rounding method used when the reference image is predicted as a target image.

The moving picture predictive decoding apparatus according to one aspect of the present invention is obtained by encoding a plurality of images constituting a moving picture by either intra prediction or inter prediction using motion compensation. , Input means for inputting compressed image data including a motion vector, restoration means for restoring a reproduced image by decoding the compressed image data, and use of the restored reproduced image for decoding subsequent images And a method of rounding a sample value generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction. and determining the rounding method determination means comprises a, the rounding method determination means, separately for the case where the case and bidirectional prediction for performing unidirectional prediction, the difference interpolated in the reference image It said rounding method to be performed in the pull position, the reference image and determines based on the information on how the rounding is used when the predicted as the target image. The moving picture predictive decoding apparatus according to one aspect of the present invention is obtained by encoding a plurality of images constituting a moving picture by either intra prediction or inter prediction using motion compensation. , Input means for inputting compressed image data including a motion vector, restoration means for restoring a reproduced image by decoding the compressed image data, and use of the restored reproduced image for decoding subsequent images And a method of rounding a sample value generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction. Rounding method determining means for determining, and the rounding method determining means individually for reference images used for two types of prediction signals used when bi-directional prediction is performed, It said rounding method to be performed at the sample position interpolated in irradiation image, the reference image and determines based on the information on how the rounding is used when the predicted as the target image.

The rounding method determination means of the moving picture predictive decoding apparatus may determine the rounding method individually for each reference image .

  Further, the rounding method determining means may determine the rounding method for at least one of the luminance signal and the color difference signal. Further, the rounding method determining means may determine the rounding method individually for each of the luminance signal and the color difference signal.

A moving image predictive encoding method according to one aspect of the present invention is a moving image executed by a moving image predictive encoding device including an image storage unit for storing a reference image used for encoding a subsequent image. An image predictive encoding method, wherein an input step for inputting a plurality of images constituting a moving image, and an input image is encoded by either intra prediction or inter prediction using motion compensation. An encoding step for generating compressed image data, a restoration step for restoring the reproduced image by decoding the generated compressed image data, and encoding the subsequent image of the restored reproduced image. An image storage step for storing the reference image used in the image storage means, and encoding the input image by the inter-screen prediction, which is necessary for fractional accuracy prediction. Comprising a rounded method determining step for determining the rounding method of the sample values generated by Motonai interpolation, the at the rounding method determination step, the moving image predictive coding apparatus, both in the case of performing unidirectional prediction The rounding method to be performed at the sample position is determined in accordance with the sample position to be interpolated in the reference image individually for the direction prediction . A video predictive encoding method according to an aspect of the present invention is executed by a video predictive encoding device including image storage means for storing a reference image used for encoding a subsequent image. An input step of inputting a plurality of images constituting a moving image, and the input image is encoded by either intra prediction or inter prediction using motion compensation. An encoding step for generating compressed image data, a restoration step for restoring the reproduced image by decoding the generated compressed image data, and encoding the subsequent image of the restored reproduced image. An image storing step of storing the reference image used as a reference image in the image storage means; and for encoding the input image by the inter-screen prediction, for the fractional accuracy prediction A rounding method determination step for determining a rounding method for sample values generated by necessary pixel interpolation, and in the rounding method determination step, the video predictive coding apparatus performs bi-directional prediction. The rounding method to be performed at the sample position is determined according to the sample position to be interpolated in the reference image individually for the reference images used for the two types of prediction signals used in the above.

A video predictive encoding method according to an aspect of the present invention is executed by a video predictive encoding device including image storage means for storing a reference image used for encoding a subsequent image. An input step of inputting a plurality of images constituting a moving image, and the input image is encoded by either intra prediction or inter prediction using motion compensation. An encoding step for generating compressed image data, a restoration step for restoring the reproduced image by decoding the generated compressed image data, and encoding the subsequent image of the restored reproduced image. An image storing step of storing the reference image used as a reference image in the image storage means; and for encoding the input image by the inter-screen prediction, for the fractional accuracy prediction Comprising a rounded method determination step to determine how rounding sample values generated by the main pixel interpolation, the at the rounding method determination step, the moving image predictive coding apparatus, when performing unidirectional prediction The rounding method to be performed individually at the sample position to be interpolated in the reference image is separately based on the information on the rounding method used when the reference image is predicted as the target image. It is characterized by determining. A video predictive encoding method according to an aspect of the present invention is executed by a video predictive encoding device including image storage means for storing a reference image used for encoding a subsequent image. An input step of inputting a plurality of images constituting a moving image, and the input image is encoded by either intra prediction or inter prediction using motion compensation. An encoding step for generating compressed image data, a restoration step for restoring the reproduced image by decoding the generated compressed image data, and encoding the subsequent image of the restored reproduced image. An image storing step of storing the reference image used as a reference image in the image storage means; and for encoding the input image by the inter-screen prediction, for the fractional accuracy prediction A rounding method determination step for determining a rounding method for sample values generated by necessary pixel interpolation, and in the rounding method determination step, the video predictive coding apparatus performs bi-directional prediction. The rounding method to be performed individually at the sample position to be interpolated in the reference image is used when the reference image is predicted as the target image individually for the reference images used for the two types of prediction signals used in The rounding method is determined based on information.

A video predictive decoding method according to one aspect of the present invention is a video predictive decoding performed by a video predictive decoding device including an image storage unit for storing a reference image used for decoding a subsequent image. A method of inputting compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation. An input step, a restoration step for restoring a reproduced image by decoding the compressed image data, and the restored reproduced image are stored in the image storage means as a reference image used for decoding a subsequent image. Sample value generated by pixel interpolation when decoding the compressed image data obtained by the image storage step and encoding by the inter-screen prediction including fractional accuracy prediction Comprising a rounded method determination step to determine how rounding, the at the rounding method determination step, the moving image prediction decoding device, separately for the case where the case and bidirectional prediction for performing unidirectional prediction, see According to a sample position to be interpolated in an image, the rounding method to be performed at the sample position is determined. The moving picture predictive decoding method according to one aspect of the present invention is a moving picture executed by a moving picture predictive decoding apparatus including an image storage unit for storing a reference image used for decoding a subsequent image. This is a predictive decoding method, and compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation. An input step for inputting, a restoration step for restoring a reproduced image by decoding the compressed image data, and the restored reproduced image as a reference image used for decoding a subsequent image in the image storage means When decoding the compressed image data obtained by the encoding by the inter-picture prediction including the image storing step to store and the fractional accuracy prediction, a sample generated by pixel interpolation is used. A rounding method determination step for determining a round value rounding method, and in the rounding method determination step, the moving picture predictive decoding device uses two types of prediction signals used when bi-directional prediction is performed. The rounding method to be performed at the sample position is determined for each reference image individually according to the sample position to be interpolated in the reference image.

The moving picture predictive decoding method according to one aspect of the present invention is a moving picture executed by a moving picture predictive decoding apparatus including an image storage unit for storing a reference image used for decoding a subsequent image. This is a predictive decoding method, and compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation. An input step for inputting, a restoration step for restoring a reproduced image by decoding the compressed image data, and the restored reproduced image as a reference image used for decoding a subsequent image in the image storage means When decoding the compressed image data obtained by the encoding by the inter-picture prediction including the image storing step to store and the fractional accuracy prediction, a sample generated by pixel interpolation is used. Comprising a rounded method determining step for determining the rounding method Le values, and in said rounding method determination step, the moving image prediction decoding device, individually with the case of performing the case and bidirectional prediction for performing unidirectional prediction In addition, the rounding method to be performed at a sample position to be interpolated in a reference image is determined based on information on a rounding method used when the reference image is predicted as a target image. The moving picture predictive decoding method according to one aspect of the present invention is a moving picture executed by a moving picture predictive decoding apparatus including an image storage unit for storing a reference image used for decoding a subsequent image. This is a predictive decoding method, and compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation. An input step for inputting, a restoration step for restoring a reproduced image by decoding the compressed image data, and the restored reproduced image as a reference image used for decoding a subsequent image in the image storage means When decoding the compressed image data obtained by the encoding by the inter-picture prediction including the image storing step to store and the fractional accuracy prediction, a sample generated by pixel interpolation is used. A rounding method determination step for determining a round value rounding method, and in the rounding method determination step, the moving picture predictive decoding device uses two types of prediction signals used when bi-directional prediction is performed. The rounding method to be performed individually on the reference image to be performed at the sample position to be interpolated in the reference image is determined based on information on the rounding method used when the reference image is predicted as the target image. It is characterized by that.

A moving image predictive encoding program according to an aspect of the present invention provides a computer, an input unit that inputs a plurality of images constituting a moving image, and an input image between screens using intra prediction or motion compensation. Encoding means for generating compressed image data by encoding by one of predictions, restoration means for restoring a reproduced image by decoding the generated compressed image data, and the restored reproduced image An image storage means for storing the subsequent image as a reference image used for encoding, and pixel interpolation necessary for fractional accuracy prediction when the input image is encoded by the inter-screen prediction. rounding method determination means for determining the rounding method of the sample values generated, to operate as the rounding method determination means, when performing case and bidirectional prediction for performing unidirectional prediction In individually, depending on the sample position interpolated in the reference image, and determines the rounding method to be performed in the sample position. The moving image predictive encoding program according to one aspect of the present invention uses a computer, input means for inputting a plurality of images constituting a moving image, and the input image using intra prediction or motion compensation. Encoding means for generating compressed image data by encoding by inter-screen prediction, restoration means for restoring a reproduced image by decoding the generated compressed image data, and the restored reproduction Image storage means for storing an image as a reference image used for encoding a subsequent image, and in-pixel necessary for fractional accuracy prediction when the input image is encoded by the inter-screen prediction It operates as a rounding method determining means for determining a rounding method for sample values generated by the interpolation, and the rounding method determining means is used in two types when performing bidirectional prediction. Individually for the reference image to be used for the prediction signal, depending on the sample position interpolated in the reference image, and determines the rounding method to be performed in the sample position.

The moving image predictive encoding program according to one aspect of the present invention uses a computer, input means for inputting a plurality of images constituting a moving image, and the input image using intra prediction or motion compensation. Encoding means for generating compressed image data by encoding by inter-screen prediction, restoration means for restoring a reproduced image by decoding the generated compressed image data, and the restored reproduction Image storage means for storing an image as a reference image used for encoding a subsequent image, and in-pixel necessary for fractional accuracy prediction when the input image is encoded by the inter-screen prediction rounding method determination means for determining the rounding method of the sample values generated by interpolation, is operated as the rounding method determination means, the line where the bidirectional prediction which performs unidirectional prediction Separately in the case, the rounding method to be performed at the sample position interpolated in the reference image, wherein the reference image is determined based on the information on how the rounding is used when the predicted as the target image And The moving image predictive encoding program according to one aspect of the present invention uses a computer, input means for inputting a plurality of images constituting a moving image, and the input image using intra prediction or motion compensation. Encoding means for generating compressed image data by encoding by inter-screen prediction, restoration means for restoring a reproduced image by decoding the generated compressed image data, and the restored reproduction Image storage means for storing an image as a reference image used for encoding a subsequent image, and in-pixel necessary for fractional accuracy prediction when the input image is encoded by the inter-screen prediction It operates as a rounding method determining means for determining a rounding method for sample values generated by the interpolation, and the rounding method determining means is used in two types when performing bidirectional prediction. The rounding method to be performed individually at the sample position to be interpolated in the reference image individually for the reference image used for the prediction signal, and information on the rounding method used when the reference image is predicted as the target image It determines based on.

A moving image predictive decoding program according to an aspect of the present invention is obtained by encoding a plurality of images constituting a moving image by using a computer by either intra prediction or inter prediction using motion compensation. In addition, input means for inputting compressed image data including a motion vector, restoration means for restoring a reproduced image by decoding the compressed image data, and decoding the restored reproduced image for subsequent images Image storage means for storing as a reference image to be used, and a method for rounding sample values generated by pixel interpolation when decoding the compressed image data obtained by encoding by inter-screen prediction including fractional accuracy prediction rounding method determination means for determining a to operate as the rounding method determination means, separately for the case where the case and bidirectional prediction for performing unidirectional prediction, reference image Depending on the sample position interpolated to definitive, and determines the rounding method to be performed in the sample position. The moving picture predictive decoding program according to one aspect of the present invention encodes a plurality of images constituting a moving picture by using either intra prediction or inter prediction using motion compensation. The obtained input means for inputting the compressed image data including the motion vector, the restoration means for restoring the reproduced image by decoding the compressed image data, and the subsequent image of the restored reproduced image are decoded. When decoding the compressed image data obtained by encoding by the inter-frame prediction including fractional accuracy prediction, the image storage means for storing the reference value used as a reference image, and the sample value generated by pixel interpolation It operates as a rounding method determination means for determining a rounding method, and the rounding method determination means is used for two types of prediction signals used in bidirectional prediction. Individually for the reference image that, depending on the sample position interpolated in the reference image, and determines the rounding method to be performed in the sample position.

The moving picture predictive decoding program according to one aspect of the present invention encodes a plurality of images constituting a moving picture by using either intra prediction or inter prediction using motion compensation. The obtained input means for inputting the compressed image data including the motion vector, the restoration means for restoring the reproduced image by decoding the compressed image data, and the subsequent image of the restored reproduced image are decoded. When decoding the compressed image data obtained by encoding by the inter-frame prediction including fractional accuracy prediction, the image storage means for storing the reference value used as a reference image, and the sample value generated by pixel interpolation rounding method determination means for determining how rounding, is operated as the rounding method determination means, separately for the case where the case and bidirectional prediction for performing unidirectional prediction, ginseng It said rounding method to be performed at the sample position interpolated in the image, the reference image and determines based on the information on how the rounding is used when the predicted as the target image. The moving picture predictive decoding program according to one aspect of the present invention encodes a plurality of images constituting a moving picture by using either intra prediction or inter prediction using motion compensation. The obtained input means for inputting the compressed image data including the motion vector, the restoration means for restoring the reproduced image by decoding the compressed image data, and the subsequent image of the restored reproduced image are decoded. When decoding the compressed image data obtained by encoding by the inter-frame prediction including fractional accuracy prediction, the image storage means for storing the reference value used as a reference image, and the sample value generated by pixel interpolation It operates as a rounding method determination means for determining a rounding method, and the rounding method determination means is used for two types of prediction signals used in bidirectional prediction. Individually determining the rounding method to be performed at the sample position to be interpolated in the reference image based on information on the rounding method used when the reference image is predicted as the target image. It is characterized by.

  According to the present invention as described above, accumulation of rounding errors is suppressed, and the reproduction image quality is improved, the prediction efficiency of an image using the reproduction image as a reference image is improved, and the subjective performance is improved.

  According to the present invention, motion compensation of fractional precision motion is used, and a sample value at a position indicated by a motion vector is generated by a pixel interpolation method. At that time, in consideration of fluctuations in the motion vector, the rounding method is determined according to the fractional pixel position to prevent accumulation of rounding errors in the entire screen, and when motion compensation using different reference images is performed. In addition, accumulation of rounding errors in the entire screen is suppressed, and the reproduction image quality is improved, the prediction efficiency of an image using the reproduction image as a reference image is improved, and the subjective performance is improved.

  Further, according to the present invention, the rounding method determination means determines the rounding method to be performed at the sample position to be interpolated in the reference image, based on information on the rounding method used when the reference image is predicted as the target image. This makes it possible to more effectively suppress the accumulation of rounding errors.

It is a figure for demonstrating the subject of a prior art. It is a functional block diagram which shows the structure of the moving image predictive coding apparatus which concerns on 1st, 2nd embodiment. It is a functional block diagram which shows the structure of the inter-screen predictor in the moving image predictive coding apparatus which concerns on 1st Embodiment. It is a functional block diagram which shows the structure of the moving image predictive decoding apparatus which concerns on 1st, 2nd embodiment. It is a functional block diagram which shows the structure of the inter-screen predictor in the moving image predictive decoding apparatus which concerns on 1st Embodiment. It is a figure for demonstrating an example of an interpolation system. It is a figure which shows the rounding method according to a position in case the precision of a motion is 1/8 pixel precision. It is a figure which shows the 1st modification of the rounding method according to a position. It is a flowchart which shows operation | movement of the inter-screen predictor in the moving image predictive coding apparatus which concerns on 1st Embodiment. It is a flowchart which shows operation | movement of the inter-screen predictor in the moving image predictive decoding apparatus which concerns on 1st Embodiment. It is a functional block diagram which shows the structure of the inter-screen predictor in the moving image predictive coding apparatus which concerns on 2nd Embodiment. It is a functional block diagram which shows the structure of the inter-screen predictor in the moving image prediction decoding apparatus which concerns on 2nd Embodiment. It is a flowchart which shows operation | movement of the inter-screen predictor which concerns on 2nd Embodiment. It is a figure which shows the 2nd modification of the rounding method according to a position. It is a figure which shows the 3rd modification of the rounding method according to a position. It is a figure which shows the 4th modification of the rounding method according to a position. It is a figure which shows the hardware constitutions of the computer for performing the program recorded on the recording medium. It is a general-view figure of the computer for performing the program memorize | stored in the recording medium. It is a block diagram which shows the structural example of a moving image predictive encoding program. It is a block diagram which shows the structural example of a moving image prediction decoding program.

  Embodiments according to the present invention will be described below with reference to FIGS.

(First embodiment)
FIG. 2 is a block diagram showing the configuration of the video predictive coding device 200 according to the first embodiment. As shown in FIG. 2, the moving image predictive encoding device 200 includes an input terminal 201, a block divider 202, a prediction signal generator 203, a frame memory 204, a subtractor 205, a converter 206, a quantum, as a functional configuration. A 207, an inverse quantizer 208, an inverse transformer 209, an adder 210, an entropy encoder 211, and an output terminal 212. The operation of each functional block will be described in the operation of the moving picture predictive coding apparatus 200 described later. The converter 206 and the quantizer 207 correspond to encoding means. The inverse quantizer 208, the inverse transformer 209, and the adder 210 correspond to decoding means.

  The operation of the moving picture predictive coding apparatus 200 configured as described above will be described below. A moving image signal composed of a plurality of images to be encoded is input to an input terminal 201, and each image is divided into a plurality of regions by a block divider 202. In the present embodiment, each image is divided into a plurality of blocks each having 8 × 8 pixels, but may be divided into other block sizes or block shapes. Next, the prediction signal generator 203 generates a prediction signal by using a prediction method to be described later for a block to be encoded (hereinafter referred to as “target block”). In the present embodiment, two types of prediction methods, inter-screen prediction and intra-screen prediction, can be used as prediction methods, and bidirectional inter-screen prediction can also be used in inter-screen prediction.

  Details of the inter-screen predictor 221 in the prediction signal generator 203 will be described with reference to FIG. 3 includes a rounding method determiner 301, a reference image interpolator 302, and a motion detector / compensator 303.

  In inter-screen prediction, first, a plurality of reproduced images restored after being encoded in the past via line L204 are inter-screen predictors 221 (rounding method determiner 301 and reference image interpolator 302). Is input. Here, the rounding method determiner 301 determines a rounding method to be used when the input reference image is interpolated. Details of the rounding method will be described later.

  Next, the reference image interpolator 302 performs pixel interpolation on the reference image according to the accuracy of motion detection in accordance with the rounding method information input from the line L301, and the reference image after the interpolation is motion detection / motion. It is sent to the compensator 303. The motion detector / compensator 303 obtains a prediction signal with the smallest error with respect to the target block from a plurality of interpolated reference images input via the line L302, and obtains motion information that is a displacement of the prediction signal. This process is called motion detection. Further, according to circumstances, the target block may be subdivided, and the inter-screen prediction method may be determined for the subdivided small area. In this case, among the various division methods, the division method of the small area and the motion information of each small area that are most efficient for the entire target block are determined. The details of the motion detection / compensation measurement method are described in the conventional techniques MPEG-2, 4, H.264. It is the same as any of the H.264 methods. Note that the motion detection / motion compensator 303 generates an inter-screen prediction signal from the reference image, the small region dividing method, and the motion information of each small region.

  The subregion division method information and the motion information of each subregion determined in this way are sent to the entropy encoder 211 via the line L212 in FIG. 2 and encoded by the entropy encoder 211. The encoded data is sent from the output terminal 212 via the line L211. Also, information (reference index) on which reference image the prediction signal is acquired from among the plurality of reference images is also sent from the prediction signal generator 203 to the entropy encoder 211 via the line L212, and the entropy encoder The encoded data is sent from the output terminal 212 via the line L211 after being encoded by the 211. In the present embodiment, four to five reproduced images are stored in the frame memory 204 and used as reference images.

  On the other hand, the intra-screen predictor 222 generates an intra-screen prediction signal using the already reproduced pixel values spatially adjacent to the target block as intra-screen prediction. Specifically, the intra-screen predictor 222 obtains an already reproduced pixel signal in the same screen from the frame memory 204, and generates an intra-screen prediction signal by extrapolating the already reproduced pixel signal. Information relating to the extrapolation method in the intra prediction is sent to the entropy encoder 211 via the line L212, encoded by the entropy encoder 211, and the encoded data is output via the line L211. It is sent out from the terminal 212.

  The basic operations of inter-screen prediction and intra-screen prediction have been outlined above. Actually, the prediction signal generator 203 selects, for each target block, the signal having the smallest error among the inter-screen prediction signal and the intra-screen prediction signal generated as described above, and the selected prediction signal is And sent to the subtracter 205 and the adder 210 via the line L203.

  The subtracter 205 subtracts the prediction signal received via the line L103 from the signal of the target block received via the line L102, and generates a residual signal. This residual signal is subjected to discrete cosine transform by a transformer 206, and each transform coefficient is quantized by a quantizer 207. Finally, the quantized transform coefficient is encoded by the entropy encoder 211, and the obtained encoded data is sent from the output terminal 212 via the line L111 together with information on the prediction method.

  On the other hand, in order to perform intra prediction or inter prediction for the subsequent target block, the quantized transform coefficient (the encoded data of the target block) is dequantized by the inverse quantizer 208. Inverse discrete cosine transform is performed by the inverse transformer 209, whereby the residual signal is restored. Then, the adder 210 adds the restored residual signal and the prediction signal sent from the line L 203, thereby reproducing the signal of the target block and storing the reproduction signal in the frame memory 204. In this embodiment, the converter 206 and the inverse converter 209 are used, but other conversion processes instead of these may be used. In some cases, the converter 206 and the inverse converter 209 may be omitted.

  Next, the video predictive decoding device according to the present embodiment will be described. FIG. 4 shows functional blocks showing the configuration of the image predictive decoding apparatus 400 according to this embodiment. As shown in FIG. 4, the moving picture predictive decoding apparatus 400 includes an input terminal 401, a data analyzer 402, an inverse quantizer 403, an inverse transformer 404, an adder 405, and a prediction signal generator 408 as functional configurations. A frame memory 407 and an output terminal 406. The operation of each functional block will be described in the operation of the moving picture predictive decoding apparatus 400 described later. Note that the decoding means is not limited to the inverse quantizer 403 and the inverse transformer 404, and other means may be used. Further, the means for decoding may be configured only by the inverse quantizer 403 without the inverse transformer 404. The inverse quantizer 403 and the inverse transformer 404 correspond to decoding means.

  The operation of the moving picture predictive decoding apparatus 400 configured as described above will be described below. The compressed data obtained by the encoding method described above is input from the input terminal 401. This compressed data includes information related to the generation of the residual signal, quantization parameter, and prediction signal of the target block. As information related to generation of a prediction signal, for example, in the case of inter-screen prediction, information related to block division (subregion division method information (eg, block size)), motion information of each small region, and the above-described reference index In the case of intra prediction, information on an extrapolation method from neighboring reproduced pixels is included.

  The data analyzer 402 extracts the residual signal of the target block, information related to the generation of the prediction signal, and the quantization parameter from the input compressed data. Among these, the residual signal and quantization parameter of the target block are sent to the inverse quantizer 403 via the line L402a, and the inverse quantizer 403 inverts the residual signal of the target block based on the quantization parameter. After the quantization, the inverse transformer 404 performs an inverse discrete cosine transform on the result of the inverse quantization. The residual signal restored in this way is sent to the adder 405 via the line L404.

  On the other hand, information regarding the generation of the extracted prediction signal is sent to the prediction signal generator 408 via the line L402b. The prediction signal generator 408 obtains an appropriate reference image from a plurality of reference images in the frame memory 407 based on information related to the generation of the prediction signal, and generates a prediction signal based on the appropriate reference image. Generate. The prediction signal generator 408 includes an inter-screen predictor 409 and an intra-screen predictor 410. 5 includes a rounding method determiner 501, a reference image interpolator 502, and a prediction signal generator 503. Details of these operations will be described later. Here, the generated prediction signal is sent to the adder 405 via the line L308a, and is added to the restored residual signal by the adder 405. As a result, the signal of the target block is reproduced. The reproduced signal of the target block is output from the output terminal 206 via the line L405 and stored in the frame memory 407 as a reproduced image.

  The rounding method used in the moving picture predictive encoding method and the moving picture predictive decoding method in this embodiment and the interpolation process of a reference image will be described below. The operation of the rounding method determiner 301 will be described below, but the operation of the rounding method determiner 501 is the same.

ここで丸め込み方法決定器３０１は、丸め込み方法として内挿されたサンプル値をサンプル位置に基づき、算出されたサンプル値の小数部分の値（以下「小数値」という）が0.5の場合に当該小数値を切り上げるか、切捨てるかを決定する。図７において輝度情報に対する動きの精度が１／８画素精度の場合の色差信号に対する丸め込み方法の位置による違いを示す。ここで、輝度情報に対する動きの精度が１／８画素精度であるため、色差信号に対する精度は１／１６画素精度であることを意味する。図７において１／１６画素精度の垂直方向位置をｋ、水平方向位置をｌとしており、丸め込み方法決定器３０１は、分数精度動きが示す位置(ｋ,l)に応じて丸め込み方法を決定する。図７においてAは丸め込み方法としてサンプル値の小数値が0.5の場合に切り上げを実施する位置を、Bは丸め込み方法としてサンプル値の小数値が0.5の場合に切り捨てを実施する位置を、それぞれ意味する。このときの決定に関する制御は、上記の式（１）におけるroundを変更することで実現される。すなわち画素位置Aではround=d²/2と設定し、画素位置Bではround=(d²/2)-1と設定する。また、図７におけるroundの値は以下のように表現される。

ここで、int(x)はxの整数部分の値を示し、＆はビット演算を示す。すなわち、

は、

が奇数の場合には、１となり、

が偶数の場合には、０となり、上記の決定に関する制御を実現することができる。 H. In motion detection / compensation used in H.264 or the like, the accuracy of motion is 1/2 pixel accuracy or 1/4 pixel accuracy in order to further improve the accuracy of the search signal. Further, motion information with 1/8 pixel accuracy with higher accuracy may be used. However, there are no sample values corresponding to these fractional pixel positions. Therefore, sample values at these fractional pixel positions are often generated by an interpolation method. Further, as a video encoded in H.264, a video format in which a color difference signal has only half the resolution with respect to luminance information is also targeted. At this time, generally, the motion information for the color difference information is defined as having half the motion information for the luminance information. In FIG. 6, assuming that the sample values of the pixels 601, 602, 603, and 604 are Ia, Ib, Ic, and Id (Ia, Ib, Ic, and Id are non-negative integers), the sample values at the positions to be interpolated ( 6 is generated by the following formula (1) using Ia, Ib, Ic, and Id.

Here, the rounding method determiner 301 uses the sample value interpolated as the rounding method based on the sample position, and when the value of the decimal part of the calculated sample value (hereinafter referred to as “decimal value”) is 0.5, the decimal value is calculated. Decide whether to round up or down. FIG. 7 shows the difference depending on the position of the rounding method with respect to the color difference signal when the accuracy of motion with respect to the luminance information is 1/8 pixel accuracy. Here, since the accuracy of the motion with respect to the luminance information is 1/8 pixel accuracy, it means that the accuracy with respect to the color difference signal is 1/16 pixel accuracy. In FIG. 7, the vertical position with 1/16 pixel accuracy is k, and the horizontal position is l. The rounding method determination unit 301 determines the rounding method according to the position (k, l) indicated by the fractional precision motion. In FIG. 7, A means a position where rounding is performed when the decimal value of the sample value is 0.5, and B means a position where rounding is performed when the decimal value of the sample value is 0.5. . Control relating to the determination at this time is realized by changing round in the above-described equation (1). That set the pixel position A in round = d ^2/2, it sets a pixel position in ^{B round = (d 2/2} ) -1. Further, the value of round in FIG. 7 is expressed as follows.

Here, int (x) indicates the value of the integer part of x, and & indicates a bit operation. That is,

Is

Is 1 if it is odd,

Is an even number, it becomes 0, and the control related to the above determination can be realized.

  In the present embodiment, the round expression is set as described above as control related to the determination of the rounding method, but the present invention is not limited to this. For example, the value of round may be determined by referring to a predetermined table using the position (k, l) as a key.

  In this embodiment, the rounding methods A and B are changed to a checkered pattern as shown in FIG. 7, but the present invention is not limited to this. As shown in FIG. 8, the positions of the rounding methods A and B in FIG. 7 may be reversed. Further, the rounding method may be changed for each column as shown in FIGS. Further, the rounding methods A and B may be set to be changed on a checkered pattern finer than that in FIG. 7 as shown in FIG. In the present embodiment, an example of 1/16 pixel accuracy is shown, but it goes without saying that the present invention can be applied to other 1 / d pixel accuracy (d is an integer of 2 or more).

  In this embodiment, the rounding method based on the same fractional pixel position is determined for all reference images, but the present invention is not limited to this. The rounding method based on different fractional pixel positions for each reference image may be changed.

  In the present embodiment, the switching of the rounding method when the fractional part of the interpolated sample value is 0.5 is shown, but the present invention is not limited to this. The method of always rounding up or discarding the interpolated sample value may be switched.

  In addition, the change of the rounding method based on the position in the present embodiment is constantly used regardless of the encoding status, but is not limited thereto. For example, the rounding method may be changed based on fractional pixel positions that differ between when performing unidirectional prediction and when performing bidirectional prediction. Further, different rounding methods may be used for reference images used for two types of prediction signals used in bidirectional prediction.

  In this embodiment, an example of a color difference signal is shown. However, the pixel value in this embodiment can be applied to a luminance signal and a color difference signal. Further, a rounding method based on a common fractional pixel position for the luminance signal and the color difference signal may be used, or a rounding method based on a different fractional pixel position may be used. Further, a rounding method based on the fractional pixel position may be determined for either the luminance signal or the color difference signal.

  Next, the operation of the reference image interpolator 302 will be described (the operation of the reference image interpolator 502 is also the same). The reference image interpolator 302 in this embodiment performs an interpolation process on the reference image using a rounding method based on the fractional pixel position determined by the rounding method determining unit 301. Specifically, the four adjacent pixels 601 to 604 are selected from the reference image as shown in FIG. 6, and the above-described interpolation method and rounding method are used according to the fractional pixel accuracy used for motion detection. Interpolation processing of sample values from four pixels is performed. A reference image in which a sample value with a predetermined fractional pixel accuracy is interpolated is used for motion detection. In the present embodiment, the bi-linear interpolation defined by the above-described equation (1) is performed.

  In the present embodiment, an example in which an interpolation method is set in advance has been described. However, the present invention is not limited to this, and information on the interpolation method may be input from the outside. In that case, information indicating the interpolation method may be transmitted from the reference image interpolator 302 to the reference image interpolator 502 so that the reference image interpolators 302 and 502 can use the same method.

  Further, although the bilinear interpolation method is shown as the interpolation method in the present embodiment, the present invention is not limited to this. A sample value may be interpolated by performing a one-dimensional interpolation formula a plurality of times.

  In this embodiment, an example in which the number of pixels used for sample interpolation is four pixels is shown, but the present invention is not limited to this. Interpolation using more pixels or only two pixels may be performed.

  In the present embodiment, the interpolation process for the reference image is performed on the entire reference image, but the present invention is not limited to this. For example, the interpolation processing may be performed for each region in the motion detection search range. Further, in the video predictive decoding apparatus, it is only necessary to determine a rounding method for the sample value at the position indicated by the fractional pixel motion vector. By calculating such sample values, interpolation processing in a limited pixel region is performed, so that an effect of reducing the amount of memory used can be obtained.

  The motion detector / compensator 303 performs motion detection on each symmetric block using a plurality of interpolated reference images, and generates an inter-screen prediction signal by motion compensation. The generated inter-screen prediction signal is sent to the differentiator 205 via the line L203. Also, information relating to block division, which is information relating to prediction signal generation determined by the motion detector / compensator 303, a motion vector, a reference index, and the like are output to the entropy encoder 211 via a line L212. The operation of the motion detector / compensator 303 is the same as that of the conventional H.264 standard. The method may be the same as that used in H.264 or MPEG4.

  On the other hand, the inter-screen prediction signal generator 503 generates and outputs a prediction signal (inter-screen prediction signal) based on information related to the prediction signal. The operation of the inter-screen prediction signal generator 503 is the same as that of the conventional H.264. The method may be the same as that used in H.264 or MPEG4.

  FIG. 9 shows a flowchart of the operation of the inter-picture predictor 221 in the video predictive coding method according to this embodiment. First, in step S <b> 901, a plurality of reference images stored in the frame memory 204 are input to the inter-screen predictor 221. Next, in step S902, the rounding method determiner 301 determines a rounding method when the fractional value of the interpolation sample value is 0.5 according to the fractional pixel position, based on the fractional pixel accuracy of the motion. The detailed determination method is as described above. In step 903, the reference image interpolator 302 performs a reference image interpolation process based on a rounding method based on the fractional pixel position determined in step 902 and a preset interpolation method. In step S904, the motion detector / compensator 303 performs motion detection / motion compensation and generates an inter-screen prediction signal. In step S905, the motion detector / compensator 303 outputs the generated inter-screen prediction signal and information related to prediction, and ends the process.

  FIG. 10 shows a flowchart of the operation of the inter-screen predictor 409 in the video predictive decoding method according to this embodiment. First, a motion vector and a reference index, which are motion information decoded in step S1001, and a reference image stored in the frame memory 407 are input to the inter-screen predictor 409. In step S1002, the rounding method determiner 501 determines a rounding method when the decimal value of the interpolation sample value is 0.5 according to the fractional pixel position indicated by the motion vector. In step S1003, the reference image interpolator 502 performs an interpolation process on the reference image based on the rounding method and interpolation method determined in step S1002. Then, in step S1004, the prediction signal generator 503 generates an inter-screen prediction signal based on the information related to the prediction signal, and in step S1005, outputs the generated inter-screen prediction signal and ends the process.

  According to the present embodiment, motion compensation of fractional precision motion is used, and a sample value at a position indicated by a fractional precision motion vector is generated by a pixel interpolation method. At that time, considering the fluctuation of the motion vector, by determining the rounding method according to the fractional pixel position indicated by the motion, the rounding method when the value less than the decimal of the sample value that causes accumulation of rounding error is 0.5 Is randomized. As a result, even when motion compensation using different reference images is performed, accumulation of rounding errors in the entire screen is suppressed, the reproduction image quality is improved, the prediction efficiency of the entire image using the reproduction image as a reference image is improved, and An improvement in subjective performance can be realized.

(Second Embodiment)
The second embodiment according to the present invention will be described below. The overall configuration of the video predictive encoding device and video predictive decoding device in the second embodiment is the same as the overall configuration of the video predictive encoding device 200 of FIG. 2 and the video predictive decoding device 400 of FIG. 4 described above. Therefore, explanation is omitted. However, the configurations and operations of the inter-screen predictor 221 in the video predictive encoding device 200 and the inter-screen predictor 409 in the video predictive decoding device 400 are different from those in the first embodiment, and will be described below. .

  The inter-screen predictor 221 in FIG. 11 includes a rounding method determiner 1101, a reference image interpolator 1102, and a motion detector / compensator 1103. Here, the reference image interpolator 1102 performs the same function as the reference image interpolator 302 of FIG. 3, and the motion detector / motion compensator 1103 performs the same function as the motion detector / motion compensator 303. Further, the rounding method determiner 1101 can input information related to prediction signal generation from the motion detection / motion compensator 1103 via the line L1103.

  12 includes a rounding method determiner 1201, a reference image interpolator 1202, and a prediction signal generator 1203. Here, the reference image interpolator 1202 performs the same function as the reference image interpolator 502 in FIG. 5, and the prediction signal generator 1203 performs the same function as the prediction signal generator 503. Further, the rounding method determiner 1201 can input information about the extracted prediction signal from the data analyzer 402 via the line L402b.

  The rounding method determiners 1101 and 1201 in the second embodiment determine a rounding method different from the rounding method used when the reference image is predicted as the target image according to the position of the sample value to be interpolated. .

  In the present embodiment, the rounding method when the decimal value of the sample value based on the fractional pixel position becomes 0.5 uses two methods, the method shown in FIG. 7 and the method shown in FIG. Here, the rounding method shown in FIG. 7 is defined as round_M1, and the rounding method shown in FIG. 8 is defined as round_M2.

  In the present embodiment, the process flow of the inter-screen predictors 221 and 409 is the same as the process flow of FIGS. 9 and 10 described above, and thus the description thereof is omitted. Therefore, hereinafter, the operation of the rounding method determiner 1101 in the inter-screen predictor 221 will be described using the flowchart of FIG. 13 (the same applies to the operation of the rounding method determiner 1201 in the inter-screen predictor 409). . As a premise, the rounding method determiner 1101 provides information on the rounding method when the reference image (reference image used for the prediction signal) determined by the motion detector / motion compensator 1103 is generated via the line L1103. Captured from the motion compensator 1103 and stored.

  The rounding method of the target image is determined by the number of times the rounding method used at the time of generating the target block is adopted as follows. The number of rounds used for the target block is counted by a counter Round_count.

  First, in step S1301, motion information and reference indexes, which are information related to prediction of each prediction block, and target block generation method information (intra-screen prediction, inter-screen prediction, etc.) are input. Next, the rounding method determiner 1101 determines in step S1302 whether the target block generation method is intra-screen prediction or inter-screen prediction. When the target block generation method is intra prediction, the rounding method determiner 1101 determines that the rounding method used for the prediction block is a predetermined rounding method (here, round_M1 as an example), Round_count [ 1 is added to round_M1] (step S1303). If the image is generated by inter-screen prediction, the rounding method determiner 1101 determines whether the image is a bidirectional prediction block (step S1304). At this time, in the case of a block generated by unidirectional prediction, the rounding method determiner 1101 adds 1 to the counter Round_count [ref_round] of the rounding method ref_round (round_M1 or round_M2) used for the reference image at the time of prediction. (Step S1305). That is, 1 is added to Round_count [round_M1] when the rounding method used for the reference image is round_M1, and 1 is added to Round_count [round_M2] when the rounding method used for the reference image is round_M2.

  On the other hand, in the case of bidirectional prediction, the rounding method determiner 1101 uses round [LIST0] as the rounding method used for the past reference image for the prediction in each direction in step S1306, and uses it for the future reference image. If the rounding method is defined as round [LIST1], it is determined whether round [LIST0] and round [LIST1] are the same rounding method. If round [LIST0] and round [LIST1] are the same rounding method as a result of the determination, 1 is added to the counter Round_count [ref_round] of the same rounding method ref_round (round_M1 or round_M2). That is, when the same rounding method is round_M1, 1 is added to Round_count [round_M1], and when the same rounding method is round_M2, 1 is added to Round_count [round_M2].

  When round [LIST0] and round [LIST1] are different, rounding method determiner 1101 adds 1 to a counter Round_count [round_M1] of a predetermined rounding method (here, round_M1 as an example). (Step S1308). Then, it is determined whether or not the predicted block is the final block of the predicted image (step S1309). Here, if it is not the final block, steps S1301 to S1308 are repeated.

  On the other hand, if it is the last block in step S1309, the rounding method determiner 1101 uses the rounding method ref_round with the larger counter Round_count in step S1310 as the rounding method used for the target image (hereinafter “representative rounding”). Method)) and save.

  Thereafter, when the target image in which the information on the representative rounding method is stored is input as a reference image at the time of the next prediction signal generation, the rounding method determiner 1101 stores the representative rounding stored on the reference image. Determine the rounding method different from the method setting. Specifically, for example, when the rounding method round_M1 is determined as the representative rounding method when the nth frame is generated as the target image, the (n + x) th frame (x is a positive integer) is used as the target image. When the nth frame is used as a reference image for generating a prediction signal, a rounding method round_M2 different from the rounding method round_M1 is determined for the nth frame.

  After determining the rounding method by the rounding method determiner 1101 as described above, information on the representative rounding method is transmitted to the reference image interpolator 1102. Then, interpolation processing is performed by the reference image interpolator 1102, motion detection / motion compensation is performed by the motion detection / motion compensator 1103 using the interpolated reference image, and a prediction signal is generated, Further, information regarding the prediction block and the prediction signal is output. The detailed operation is as described above. Also, information regarding the prediction signal is sent from the motion detection / motion compensator 1103 to the rounding method determiner 1101 via the line L1103 in order to determine the representative rounding method of the target image.

  In the present embodiment, the rounding method round_M1 is set when generated by intra prediction, but the present invention is not limited to this, and the rounding method round_M2 may be set. Other rounding methods may be used. At this time, the rounding method based on the fractional pixel position is desirable.

  In the present embodiment, the bi-directional prediction block generated by synthesizing the prediction signal generated from the reference image using different rounding methods is treated as the rounding method setting round_M1, but is not limited thereto. Of the past and future reference images, the rounding method setting with the higher reference image quality may be handled. In this case, the determination may be made using the quantization parameter used for the reference image. Further, if objective performance such as PSNR is held for each reference image, the value may be used.

  In addition, as a rounding method when generating the bidirectional prediction block, rounding up or rounding down may be determined with reference to the rounding method used for the prediction signal from each direction.

  In the present embodiment, an example in which there are two types of rounding methods has been described, but the present invention is not limited to this. For example, the third rounding method round_M3 may be used for intra prediction or for bi-directional prediction in which prediction signals generated by reference images using different rounding method settings are combined.

  In the present embodiment, the number of representative rounding methods is one, but the number is not limited to this. A plurality of representative rounding methods may be determined for one reference image. For example, the representative rounding method may be determined in accordance with the positions of a plurality of regions in the screen.

  In this embodiment, the representative rounding method is determined based on the block prediction mode, but the present invention is not limited to this. The representative rounding method may be determined based on other information related to prediction signal generation. For example, the representative rounding method may be determined based on the position of the reference image and the motion vector. Also, the representative rounding method may be determined based on the quantization parameter used for the reference image.

  In this embodiment, the representative rounding method is determined based on the number of times the selected rounding method is used. However, the rounding method used for the pixel at the position indicated by the motion vector may be used. A representative rounding method may be determined.

  In addition, in the case of a target image that is not used as a reference image, the above process may not be performed. As a result, it is possible to reduce the processing load of moving picture predictive encoding and decoding without extra calculations.

  In the present embodiment, as the representative rounding method used for the target image, the rounding method used most frequently in the target image is determined as the representative rounding method. However, the present invention is not limited to this. The rounding method that satisfies a certain condition and is selected most often may be determined as the representative rounding method. For example, the most frequently used rounding method is determined as the representative rounding method only when the most used rounding method exceeds Z [%] (defined here as Z = 66) among the blocks included in the image. May be. If the above condition is not satisfied, a predetermined rounding method (for example, round_M1) may be determined as the representative rounding method. In addition, a rounding method different from the most frequently used rounding method may be determined as the representative rounding method.

  In addition, if there is no bias in the setting of the reference image rounding method used when generating the target image in the above example, the representative rounding method setting of a plurality of reference images stored in the frame memory is confirmed, and there is no bias. Of the plurality of reference image representative rounding methods, a small number of representative rounding methods may be determined for the target image. For example, when three of four reference images existing in the reference image buffer are generated with the rounding method setting round_M1, when using a reference image generated without bias in the rounding method setting, A representative rounding method setting round_M2 different from the above rounding method setting round_M1 may be used.

  According to the present embodiment, when generating a prediction signal using a plurality of reference images generated by using a plurality of rounding methods, a fractional pixel is added after taking into account the rounding method used for each reference image. Since the rounding method can be determined according to the position, the randomness of the rounding method based on the fractional pixel position for each block can be further improved, and accumulation of rounding errors can be more effectively suppressed.

  In the second embodiment, the rounding method is determined according to the fractional pixel position in consideration of the rounding method used for each reference image. However, when generating a prediction signal using a plurality of reference images, The rounding method may be determined in consideration of the rounding method used for each reference image without performing the rounding method according to the fractional pixel position. In this way, even if the rounding method used for each reference image is taken into account and only the rounding method is determined, the accumulation of rounding errors can be effectively suppressed, the reproduction image quality is improved, and the reproduction image is used as the reference image. It is possible to improve the prediction efficiency and the subjective performance of the entire image to be used.

[About a video predictive encoding program and a video predictive decoding program]
The invention relating to the moving picture predictive coding apparatus can be understood as an invention relating to a moving picture predictive coding program for causing a computer to function as a moving picture predictive coding apparatus. Similarly, the invention relating to the moving picture predictive decoding apparatus can be regarded as an invention relating to a moving picture predictive decoding program for causing a computer to function as a moving picture predictive decoding apparatus.

  The moving picture predictive encoding program and the moving picture predictive decoding program are provided by being stored in a recording medium, for example. Examples of the recording medium include a recording medium such as a flexible disk, a CD-ROM, a DVD, a recording medium such as a ROM, or a semiconductor memory.

  FIG. 19 shows modules of a moving picture predictive coding program for causing a computer to function as a moving picture predictive coding apparatus. As shown in FIG. 19, the moving picture predictive encoding program P200 includes an input module P201, an encoding module P202, a restoration module P203, an image storage module P204, and a rounding method determination module P205.

  FIG. 20 shows a module of a moving picture predictive decoding program for causing a computer to function as a moving picture predictive decoding apparatus. As shown in FIG. 20, the moving picture predictive decoding program P400 includes an input module P401, a restoration module P402, an image storage module P403, and a rounding method determination module P404.

  The moving picture predictive encoding program P200 and the moving picture predictive decoding program P400 configured as described above can be stored in the recording medium 10 shown in FIG. 18, and are executed by the computer 30 shown in FIG.

  FIG. 17 is a diagram illustrating a hardware configuration of a computer for executing a program recorded on a recording medium, and FIG. 18 is an overview of the computer for executing a program stored on the recording medium. Examples of the computer include a DVD player, a set top box, a mobile phone, etc. that have a CPU and perform processing and control by software.

  As shown in FIG. 17, the computer 30 includes a reading device 12 such as a flexible disk drive device, a CD-ROM drive device, and a DVD drive device, a working memory (RAM) 14 in which an operating system is resident, and a recording medium 10. A memory 16 for storing programs stored therein, a display device 18 such as a display, a mouse 20 and a keyboard 22 as input devices, a communication device 24 for transmitting and receiving data, and a CPU 26 for controlling execution of the programs. And. When the recording medium 10 is inserted into the reading device 12, the computer 30 can access the moving image predictive encoding program stored in the recording medium 10 from the reading device 12, and the moving image predictive encoding program can It becomes possible to operate as a moving picture predictive encoding apparatus according to the invention. Similarly, when the recording medium 10 is inserted into the reading device 12, the computer 30 can access the moving picture predictive decoding program stored in the recording medium 10 from the reading apparatus 12, and the moving picture predictive decoding program can It becomes possible to operate as a moving picture predictive decoding apparatus according to the present invention.

  As shown in FIG. 18, the moving picture predictive encoding program or the moving picture predictive decoding program may be provided via a network as a computer data signal 40 superimposed on a carrier wave. In this case, the computer 30 can store the moving picture predictive encoding program or the moving picture predictive decoding program received by the communication device 24 in the memory 16 and execute it.

  DESCRIPTION OF SYMBOLS 10 ... Recording medium, 30 ... Computer, 200 ... Moving image predictive coding apparatus, 201 ... Input terminal, 202 ... Block divider, 203 ... Prediction signal generator, 204 ... Frame memory, 205 ... Subtractor, 206 ... Converter 207: quantizer 208: inverse quantizer 209 ... inverse transformer 210 ... adder 211 ... entropy encoder 212 ... output terminal 221 inter-screen predictor 222 ... intra-screen predictor , 301 ... rounding method determiner, 302 ... reference image interpolator, 303 ... motion detection / motion compensator, 400 ... moving picture predictive decoding device, 401 ... input terminal, 402 ... data analyzer, 403 ... inverse quantizer 404: Inverter, 405 ... Adder, 406 ... Output terminal, 407 ... Frame memory, 408 ... Prediction signal generator, 409 ... Inter-screen predictor, 410 ... In-screen predictor, DESCRIPTION OF SYMBOLS 01 ... Rounding method determiner, 502 ... Reference image interpolator, 503 ... Inter-screen prediction signal generator, 1101 ... Rounding method determiner, 1102 ... Reference image interpolator, 1103 ... Motion detection / motion compensator, 1201 ... Rounding method determiner, 1202 ... reference image interpolator, 1203 ... inter-frame prediction signal generator, P200 ... moving picture predictive encoding program, P201 ... input module, P202 ... encoding module, P203 ... restoration module, P204 ... image Storage module, P205 ... rounding method determination module, P400 ... moving picture predictive decoding program, P401 ... input module, P402 ... restoration module, P403 ... image storage module, P404 ... rounding method determination module.

Claims (32)

Input means for inputting a plurality of images constituting a moving image;
Encoding means for generating compressed image data by encoding the input image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring a reproduced image by decoding the generated compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for encoding a subsequent image;
A rounding method determination means for determining a rounding method of sample values generated by pixel interpolation necessary for fractional accuracy prediction when the input image is encoded by the inter-screen prediction;
Comprising
The rounding method determining means determines the rounding method to be performed at the sample position according to the sample position to be interpolated in the reference image separately for the case of performing the unidirectional prediction and the case of performing the bidirectional prediction. ,
A video predictive coding apparatus characterized by the above. Input means for inputting a plurality of images constituting a moving image;
Encoding means for generating compressed image data by encoding the input image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring a reproduced image by decoding the generated compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for encoding a subsequent image;
A rounding method determination means for determining a rounding method of sample values generated by pixel interpolation necessary for fractional accuracy prediction when the input image is encoded by the inter-screen prediction;
Comprising
The rounding method determination means is performed individually at the sample position in accordance with the sample position to be interpolated in the reference image, with respect to the reference image used for the two types of prediction signals used when bidirectional prediction is performed. Determining the rounding method
A video predictive coding apparatus characterized by the above. The rounding method determining means determines the rounding method based on the information on the rounding method used when the reference image is predicted as a target image.
The moving picture predictive coding apparatus according to claim 1 or 2 , characterized in that Input means for inputting a plurality of images constituting a moving image;
Encoding means for generating compressed image data by encoding the input image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring a reproduced image by decoding the generated compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for encoding a subsequent image;
A rounding method determination means for determining a rounding method of sample values generated by pixel interpolation necessary for fractional accuracy prediction when the input image is encoded by the inter-screen prediction;
Comprising
The rounding method determining means predicts the rounding method to be performed at a sample position to be interpolated in a reference image , separately for one-way prediction and two-way prediction, with the reference image as a target image. Determine based on the rounding method information used when
A video predictive coding apparatus characterized by the above. Input means for inputting a plurality of images constituting a moving image;
Encoding means for generating compressed image data by encoding the input image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring a reproduced image by decoding the generated compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for encoding a subsequent image;
A rounding method determination means for determining a rounding method of sample values generated by pixel interpolation necessary for fractional accuracy prediction when the input image is encoded by the inter-screen prediction;
Comprising
The rounding method determining means is configured to perform the rounding method to be performed individually at a sample position to be interpolated in a reference image, separately for a reference image used for two types of prediction signals used in bi-directional prediction. Determine based on information of the rounding method used when the reference image is predicted as the target image,
A video predictive coding apparatus characterized by the above. The rounding method determining means determines the rounding method individually for each reference image.
Moving image prediction encoding device according to any one of claim 1 to 5, characterized in that. The rounding method determining means determines the rounding method for at least one of a luminance signal and a color difference signal.
The moving picture predictive coding apparatus according to any one of claims 1 to 6 , wherein the moving picture predictive coding apparatus is provided. The rounding method determining means determines the rounding method individually for each of the luminance signal and the color difference signal.
The moving picture predictive coding apparatus according to any one of claims 1 to 7 , wherein An input means for inputting compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring the reproduced image by decoding the compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for decoding subsequent images;
A rounding method determination means for determining a rounding method of sample values generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction;
Comprising
The rounding method determining means determines the rounding method to be performed at the sample position according to the sample position to be interpolated in the reference image separately for the case of performing the unidirectional prediction and the case of performing the bidirectional prediction. ,
A video predictive decoding apparatus characterized by the above. An input means for inputting compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring the reproduced image by decoding the compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for decoding subsequent images;
A rounding method determination means for determining a rounding method of sample values generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction;
Comprising
The rounding method determination means is performed individually at the sample position in accordance with the sample position to be interpolated in the reference image, with respect to the reference image used for the two types of prediction signals used when performing bidirectional prediction Determining the rounding method
A video predictive decoding apparatus characterized by the above. The rounding method determining means determines the rounding method based on the information on the rounding method used when the reference image is predicted as a target image.
The moving picture predictive decoding apparatus according to claim 9 or 10 , characterized in that: An input means for inputting compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring the reproduced image by decoding the compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for decoding subsequent images;
A rounding method determination means for determining a rounding method of sample values generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction;
Comprising
The rounding method determining means predicts the rounding method to be performed at a sample position to be interpolated in a reference image , separately for one-way prediction and two-way prediction, with the reference image as a target image. Determine based on the rounding method information used when
A video predictive decoding apparatus characterized by the above. An input means for inputting compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring the reproduced image by decoding the compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for decoding subsequent images;
A rounding method determination means for determining a rounding method of sample values generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction;
Comprising
The rounding method determining means is configured to perform the rounding method to be performed individually at a sample position to be interpolated in a reference image, separately for a reference image used for two types of prediction signals used in bi-directional prediction. Determine based on information of the rounding method used when the reference image is predicted as the target image,
A video predictive decoding apparatus characterized by the above. The rounding method determining means determines the rounding method individually for each reference image.
Moving image predictive decoding apparatus according to any one of claims 9 to 1 3, characterized in that. The rounding method determining means determines the rounding method for at least one of a luminance signal and a color difference signal.
Moving image predictive decoding apparatus according to any one of claims 9-1 4, characterized in that. The rounding method determining means determines the rounding method individually for each of the luminance signal and the color difference signal.
Moving image predictive decoding apparatus according to any one of claims 9 to 1 5, characterized in that. A video predictive encoding method executed by a video predictive encoding device comprising an image storage means for storing a reference image used for encoding a subsequent image,
An input step for inputting a plurality of images constituting the moving image;
An encoding step of generating compressed image data by encoding the input image by either intra-screen prediction or inter-screen prediction using motion compensation;
A restoration step of restoring a reproduced image by decoding the generated compressed image data;
An image storage step of storing the restored reproduced image in the image storage means as a reference image used for encoding a subsequent image;
A rounding method determination step for determining a rounding method of sample values generated by pixel interpolation necessary for fractional accuracy prediction when the input image is encoded by the inter-screen prediction;
Comprising
In the rounding method determination step, the moving image predictive coding apparatus separately determines the sample position according to the sample position to be interpolated in the reference image separately when performing unidirectional prediction and when performing bidirectional prediction. Determining the rounding method to be performed in
A video predictive encoding method characterized by the above. A video predictive encoding method executed by a video predictive encoding device comprising an image storage means for storing a reference image used for encoding a subsequent image,
An input step for inputting a plurality of images constituting the moving image;
An encoding step of generating compressed image data by encoding the input image by either intra-screen prediction or inter-screen prediction using motion compensation;
A restoration step of restoring a reproduced image by decoding the generated compressed image data;
An image storage step of storing the restored reproduced image in the image storage means as a reference image used for encoding a subsequent image;
A rounding method determination step for determining a rounding method of sample values generated by pixel interpolation necessary for fractional accuracy prediction when the input image is encoded by the inter-screen prediction;
Comprising
In the rounding method determination step, the moving image predictive encoding device individually interpolates sample positions in a reference image for reference images used for two types of prediction signals used when bi-directional prediction is performed. In response to determining the rounding method to be performed at the sample location,
A video predictive encoding method characterized by the above. A video predictive encoding method executed by a video predictive encoding device comprising an image storage means for storing a reference image used for encoding a subsequent image,
An input step for inputting a plurality of images constituting the moving image;
An encoding step of generating compressed image data by encoding the input image by either intra-screen prediction or inter-screen prediction using motion compensation;
A restoration step of restoring a reproduced image by decoding the generated compressed image data;
An image storage step of storing the restored reproduced image in the image storage means as a reference image used for encoding a subsequent image;
A rounding method determination step for determining a rounding method of sample values generated by pixel interpolation necessary for fractional accuracy prediction when the input image is encoded by the inter-screen prediction;
Comprising
In the rounding method determination step , the video predictive encoding apparatus is configured to perform the rounding method to be performed at a sample position to be interpolated in a reference image separately when performing unidirectional prediction and when performing bidirectional prediction. Is determined based on information on a rounding method used when the reference image is predicted as a target image.
A video predictive encoding method characterized by the above. A video predictive encoding method executed by a video predictive encoding device comprising an image storage means for storing a reference image used for encoding a subsequent image,
An input step for inputting a plurality of images constituting the moving image;
An encoding step of generating compressed image data by encoding the input image by either intra-screen prediction or inter-screen prediction using motion compensation;
A restoration step of restoring a reproduced image by decoding the generated compressed image data;
An image storage step of storing the restored reproduced image in the image storage means as a reference image used for encoding a subsequent image;
A rounding method determination step for determining a rounding method of sample values generated by pixel interpolation necessary for fractional accuracy prediction when the input image is encoded by the inter-screen prediction;
Comprising
In the rounding method determination step , the moving image predictive encoding device individually interpolates sample positions in a reference image with respect to reference images used for two types of prediction signals used when bi-directional prediction is performed. Determining the rounding method to be performed based on information on the rounding method used when the reference image is predicted as a target image,
A video predictive encoding method characterized by the above. A moving image predictive decoding method executed by a moving image predictive decoding device including an image storage means for storing a reference image used for decoding a subsequent image,
An input step of inputting compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation;
A restoration step of restoring a reproduced image by decoding the compressed image data;
An image storage step of storing the restored reproduced image in the image storage means as a reference image used for decoding a subsequent image;
A rounding method determination step for determining a rounding method for sample values generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction;
Comprising
In the rounding method determination step, the moving picture predictive decoding apparatus separately performs the unidirectional prediction and the bi-directional prediction at the sample position according to the sample position to be interpolated in the reference image. Determine the rounding method to be performed,
A video predictive decoding method characterized by the above. A moving image predictive decoding method executed by a moving image predictive decoding device including an image storage means for storing a reference image used for decoding a subsequent image,
An input step of inputting compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation;
A restoration step of restoring a reproduced image by decoding the compressed image data;
An image storage step of storing the restored reproduced image in the image storage means as a reference image used for decoding a subsequent image;
A rounding method determination step for determining a rounding method for sample values generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction;
Comprising
In the rounding method determination step, the video predictive decoding apparatus individually sets the sample position to be interpolated in the reference image with respect to the reference image used for the two types of prediction signals used when bi-directional prediction is performed. In response, determining the rounding method to be performed at the sample location,
A video predictive decoding method characterized by the above. A moving image predictive decoding method executed by a moving image predictive decoding device including an image storage means for storing a reference image used for decoding a subsequent image,
An input step of inputting compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation;
A restoration step of restoring a reproduced image by decoding the compressed image data;
An image storage step of storing the restored reproduced image in the image storage means as a reference image used for decoding a subsequent image;
A rounding method determination step for determining a rounding method for sample values generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction;
Comprising
In the rounding method determination step, the moving picture predictive decoding apparatus performs the rounding method to be performed at a sample position to be interpolated in a reference image separately when performing unidirectional prediction and when performing bidirectional prediction. , Based on the information of the rounding method used when the reference image is predicted as the target image,
A video predictive decoding method characterized by the above. A moving image predictive decoding method executed by a moving image predictive decoding device including an image storage means for storing a reference image used for decoding a subsequent image,
An input step of inputting compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation;
A restoration step of restoring a reproduced image by decoding the compressed image data;
An image storage step of storing the restored reproduced image in the image storage means as a reference image used for decoding a subsequent image;
A rounding method determination step for determining a rounding method for sample values generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction;
Comprising
In the rounding method determination step, the moving picture predictive decoding apparatus individually uses reference positions used for two types of prediction signals used when performing bi-directional prediction at sample positions to be interpolated in the reference picture. The rounding method to be performed is determined based on information on a rounding method used when the reference image is predicted as a target image.
A video predictive decoding method characterized by the above. Computer
Input means for inputting a plurality of images constituting a moving image;
Encoding means for generating compressed image data by encoding the input image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring a reproduced image by decoding the generated compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for encoding a subsequent image;
A rounding method determining means for determining a rounding method of sample values generated by pixel interpolation required for fractional accuracy prediction when the input image is encoded by the inter-screen prediction;
Act as
The rounding method determining means determines the rounding method to be performed at the sample position according to the sample position to be interpolated in the reference image separately for the case of performing the unidirectional prediction and the case of performing the bidirectional prediction. ,
A moving picture predictive encoding program characterized by the above. Computer
Input means for inputting a plurality of images constituting a moving image;
Encoding means for generating compressed image data by encoding the input image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring a reproduced image by decoding the generated compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for encoding a subsequent image;
A rounding method determining means for determining a rounding method of sample values generated by pixel interpolation required for fractional accuracy prediction when the input image is encoded by the inter-screen prediction;
Act as
The rounding method determination means is performed individually at the sample position in accordance with the sample position to be interpolated in the reference image, with respect to the reference image used for the two types of prediction signals used when bidirectional prediction is performed. Determining the rounding method
A moving picture predictive encoding program characterized by the above. Computer
Input means for inputting a plurality of images constituting a moving image;
Encoding means for generating compressed image data by encoding the input image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring a reproduced image by decoding the generated compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for encoding a subsequent image;
A rounding method determining means for determining a rounding method of sample values generated by pixel interpolation required for fractional accuracy prediction when the input image is encoded by the inter-screen prediction;
Act as
The rounding method determining means predicts the rounding method to be performed at a sample position to be interpolated in a reference image , separately for one-way prediction and two-way prediction, with the reference image as a target image. Determine based on the rounding method information used when
A moving picture predictive encoding program characterized by the above. Computer
Input means for inputting a plurality of images constituting a moving image;
Encoding means for generating compressed image data by encoding the input image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring a reproduced image by decoding the generated compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for encoding a subsequent image;
A rounding method determining means for determining a rounding method of sample values generated by pixel interpolation required for fractional accuracy prediction when the input image is encoded by the inter-screen prediction;
Act as
The rounding method determining means is configured to perform the rounding method to be performed individually at a sample position to be interpolated in a reference image, separately for a reference image used for two types of prediction signals used in bi-directional prediction. Determine based on information of the rounding method used when the reference image is predicted as the target image,
A moving picture predictive encoding program characterized by the above. Computer
An input means for inputting compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring the reproduced image by decoding the compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for decoding subsequent images;
A rounding method determination means for determining a rounding method of sample values generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction;
Act as
The rounding method determining means determines the rounding method to be performed at the sample position according to the sample position to be interpolated in the reference image separately for the case of performing the unidirectional prediction and the case of performing the bidirectional prediction. ,
A moving picture predictive decoding program characterized by the above. Computer
An input means for inputting compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring the reproduced image by decoding the compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for decoding subsequent images;
A rounding method determination means for determining a rounding method of sample values generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction;
Act as
The rounding method determination means is performed individually at the sample position in accordance with the sample position to be interpolated in the reference image, with respect to the reference image used for the two types of prediction signals used when performing bidirectional prediction Determining the rounding method
A moving picture predictive decoding program characterized by the above. Computer
An input means for inputting compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring the reproduced image by decoding the compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for decoding subsequent images;
A rounding method determination means for determining a rounding method of sample values generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction;
Act as
The rounding method determining means predicts the rounding method to be performed at a sample position to be interpolated in a reference image , separately for one-way prediction and two-way prediction, with the reference image as a target image. Determine based on the rounding method information used when
A moving picture predictive decoding program characterized by the above. Computer
An input means for inputting compressed image data including a motion vector obtained by encoding a plurality of images constituting a moving image by either intra-screen prediction or inter-screen prediction using motion compensation;
Restoring means for restoring the reproduced image by decoding the compressed image data;
Image storage means for storing the restored reproduced image as a reference image used for decoding subsequent images;
A rounding method determination means for determining a rounding method of sample values generated by pixel interpolation when decoding the compressed image data obtained by encoding by the inter-screen prediction including fractional accuracy prediction;
Act as
The rounding method determining means is configured to perform the rounding method to be performed individually at a sample position to be interpolated in a reference image, separately for a reference image used for two types of prediction signals used in bi-directional prediction. Determine based on information of the rounding method used when the reference image is predicted as the target image,
A moving picture predictive decoding program characterized by the above.

Images