JP3127956B2 - Local image decoder and local image decoding method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to, for example, an image decoder for performing decoding corresponding to an inter-frame adaptive coding system for encoding a digital moving image signal with high efficiency.

[0002]

2. Description of the Related Art FIG. 18 shows, for example, S. Nogaki, M. Ohta, T. Om
achi ■ A Study on HDTV Signal Coding with Motion Ad
2 is a block diagram showing a conventional inter-frame predictive coding method described in Aptive Noise Reduction II, (The 3rd HDTV International Workshop Preliminary Report, Vol. 3, 1989.) Is a motion detector, 3 is a differentiator, 4 is an encoder, 5 is a local decoder, 6 is an adder, and 7 is a multiplexer. Although omitted in the figure, the coded data is decoded in the same manner on the decoding side of the transmission destination to reproduce the image signal.

Next, the operation will be described with reference to FIG. With respect to the input image signal 101 that is interlaced and forms one frame from both odd and even fields, the motion of the input image signal 101 with the same field of the previous frame is detected in units of blocks in which a plurality of pixels are put together. The motion of the odd field is detected by the motion detecting unit 2 with respect to the encoding target block of the input image signal 101 in the vicinity of the position corresponding to the encoding target block of the already encoded odd field in the frame memory 1. Block 102
By searching for the most similar block from As the evaluation value of the similarity, a sum of absolute differences or a sum of squares of differences between corresponding pixels in both blocks is used. The horizontal and vertical motion amounts of the encoding target block with respect to the most similar block obtained here are output as the motion vector 103. The motion compensation prediction signal 10 corresponding to the motion vector 103 is output from the frame memory 1.
4 is output.

[0004] A prediction error signal 105 obtained by subtracting the motion compensation prediction signal 104 from the input signal 101 by the differentiator 3.
Is input to the encoding unit 4 to remove spatial redundancy. In general, a low frequency component of an image signal occupies a large component in terms of power. Therefore, information is compressed by performing quantization with a large number of bits in a high power portion and a small number of bits in a low power portion. As this method, for example, orthogonal transform such as discrete cosine transform is performed on an 8 × 8 pixel block to perform frequency transform, and transform coefficients are scalar-quantized. The scalar-quantized encoded data 106 is sent to the local decoding unit 5 and the multiplexing unit 7. The multiplexing unit 7 multiplexes the coded data 106 and the motion vector 103, performs transmission path encoding, and sends the result to the transmission path 109.

On the other hand, the local decoding unit 5 performs an operation opposite to that of the encoding unit 4, that is, inverse scalar quantization and inverse orthogonal transform, and a decoding error signal 107 is obtained. The adder 6 adds the motion-compensated prediction signal 104 to the decoded error signal 107. The obtained local decoded signal 108 is stored in the frame memory 1 and used to detect the motion of the odd field of the next frame.

Similarly, motion detection is performed on the even-numbered field of the input image signal 101 with the already-encoded field of the frame memory 1, and the motion-compensated prediction error signal is encoded. As described above, in the conventional inter-frame prediction coding method, the temporal redundancy included in the moving image signal is removed by the motion compensation prediction coding, and the orthogonal redundancy or the like is used to remove the spatial redundancy. .

[0007]

The conventional inter-frame prediction coding method predicts the current odd field from the odd field of an already coded frame, and converts the current even field from the even field of the already coded frame. Predicted and odd and even fields are individually encoded, so that the spatial correlation existing between interlaced consecutive fields is not used, so that the encoding efficiency is poor. was there.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems. For example, an encoding method for performing motion search from both odd and even fields of a frame which has already been encoded to predict each field. It is an object of the present invention to obtain an image decoder that performs decoding corresponding to a method.

[0009]

In order to solve the above-mentioned problems, according to the present invention, an interlaced input image signal composed of odd / even fields and a plurality of prediction signals are selected.
The resulting encoded data the prediction error signal is a differential motion compensated prediction is generated by coding the prediction signal, it decodes based on the motion vectors used in motion compensation prediction, a local picture decoder Decoding means for decoding encoded data of the prediction error signal to generate a decoded prediction error signal ; and a decoding prediction error decoded by the decoding means.
A signal, by adding the above prediction signal, adding means for outputting a decoded image signal, a field memory for storing divides the decoded image signal output by the adding means into odd / even fields, the field Prediction signal output means for outputting a field prediction signal of an odd / even field based on the motion vector from the decoded image signal stored in the memory; and an odd / even field output from the field memory by the prediction signal output means.
And interpolation means for generating an interpolated predictive signal by entering the field prediction signal, the odd / even field output from the field memory by the prediction signal output means Fi
And Rudo prediction signal, the prediction signal from among the interpolated predictive signal output from said interpolation means, output to said adding means
And output means for inputting power .

The interpolating means may include an odd / even number feed.
Prediction by calculating the average of the prediction signals
A signal is generated. Also, interlaced scanning
An input image signal consisting of odd / even fields and multiple
The prediction that is the difference from the prediction signal selected from the measurement signals
The encoded data generated by encoding the measurement error signal is
Based on the multiple motion vectors used in the
An image encoder having an image decoder for decoding
In addition, the image decoder performs coding data of the prediction error signal.
Decoding means for decoding the data and generating a decoded prediction error signal.
And a prediction error signal decoded by the decoding means.
And a prediction signal to be described later to generate a decoded image signal.
Adding means for inputting data, and the decoding output by the adding means.
Signal image signal is divided into odd / even fields and stored.
Field memory and the field memory
From the decoded image signal based on the plurality of motion vectors.
Prediction signal for outputting a prediction signal of odd / even fields
Output means and the prediction signal output means.
Prediction of odd / even fields output from field memory
An interpolation means for inputting a signal and generating an interpolation prediction signal;
From the field memory by the prediction signal output means.
The output prediction signal of the odd / even field and the complement
One of the interpolation prediction signals output from the
Selection means for selecting a prediction signal and outputting it to the addition means
And the following.

For this reason, in the image decoder according to the first aspect of the present invention, the already decoded image signal is divided into a plurality of fields and stored, and a plurality of prediction signals for predicting a change in the image signal based on the field signal are stored. And generating an interpolation prediction signal from at least two different prediction signals, selecting one prediction signal from the plurality of prediction signals and the interpolation prediction signal, and based on the decoded prediction error signal and the selected prediction signal. To generate and output a decoded image signal. Accordingly, even when the motion compensation prediction encoding is performed on the encoding side using a signal obtained by interpolating the prediction signals from both fields of the already encoded frame as the prediction signal, the decoding can be performed correctly. .

[0012] In the image decoder according to the next invention,
When generating an interpolation prediction signal from two different prediction signals,
This is performed by calculating an average of a plurality of prediction signals.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION

Embodiment 1 FIG. FIG. 1 is a block diagram of an adaptive field / frame coding system according to an embodiment of the present invention. Reference numeral 8 denotes an odd field memory for storing a local decoded signal of an odd field, and 9 stores a local decoded signal of an even field. An even field memory 10 is an interpolation unit that interpolates a motion compensated prediction signal from the two fields, 1
Reference numeral 1 denotes a selector for selecting a prediction signal that gives optimal prediction from a total of three prediction signals predicted from odd and even fields and the interpolated prediction signal.

Further, 200 is a motion detecting means, 300 is a prediction error signal output means, and 500 is an encoding means.

FIG. 2 is a diagram showing a state in which time is taken on the horizontal axis and the vertical direction is taken on the vertical axis of the input image signal 101 in which interlaced scanning and odd and even fields are alternately input. In FIG. 2, K1 indicates an odd field of the first frame, and G1 indicates an even frame of the first frame. Similarly, K2 indicates an odd field of the second frame, and G2 indicates an even field of the second frame.

FIG. 3 is an example of a block diagram of the interpolation unit 10. As shown in FIG. A simple averaging of the input motion compensated prediction signal 104a from the odd field and the motion compensated prediction signal 104b from the even field is taken, and the output is used as an interpolation prediction signal 104c.

Next, the operation will be described with reference to FIGS. 1, 2 and 3. With respect to the input image signal 101 in which the interlaced scanning is performed and the odd and even fields are alternately input, the motion detection with the odd and even fields of the previous frame is performed in (n × m) block units in which a plurality of pixels are combined.
In the motion detection unit 2, the motion detection unit 2 detects the position of the field to be coded in the field of the input image signal 101 corresponding to the coded field of the already coded odd field in the odd field memory 8. The search is performed by searching for the most similar block from the neighboring blocks 102a centered at.

For example, as shown in FIG. 4, in the previous frame, an image H1 exists in one block unit of (n × m), and the location of H1 is included in the image signal inputted this time. From the location of H2
The motion detector 2 outputs a motion vector 103 indicating that the block has moved from H1 to H2 in the horizontal direction. In this case, since there is no motion amount in the vertical direction, the motion vector 103 is output as 0 in the vertical direction. The horizontal and vertical motion amounts obtained in this manner are output as motion vectors 103.

The odd field memory 8 outputs a motion compensation prediction signal 104a corresponding to the motion vector 103. Similarly, motion compensation with the even field of the previous frame is performed in the motion detection unit 2 by searching for a similar block from the neighboring block 102b in the even field memory 9 for the encoding target block of the input image signal 101. And output as a motion vector 103. The even field memory 9 outputs a motion compensation prediction signal 104b corresponding to the motion vector 103.

A motion compensated prediction signal 104 motion-compensated from the odd field memory 8 according to the motion vector 103
a and the motion vector 10 from the second field memory 9.
Interpolation processing is performed by the interpolation unit 10 shown in FIG. 2 from the motion-compensated prediction signal 104b that has been motion-compensated in accordance with No. 3 to generate an interpolation prediction signal 104c. A motion compensated prediction signal 104a obtained from the odd field, a motion compensated prediction signal 104b obtained from the even field,
Among the interpolated motion-compensated prediction signals 104 c, the selector 11 selects a prediction signal having the minimum error signal power between the input image signal 101 and the encoding target block, and outputs the selected signal as the prediction signal 110.

FIG. 5 is a diagram for explaining this operation. It is assumed that the odd field memory 8 shown in FIG. 1 stores the odd field K1 of the previous frame, and the even field memory 9 of FIG. 1 stores the even field G1 of the previous frame of FIG. Here, the input image signal 101 is used as the current frame, and the odd field K
The operation when 2 and the even field G2 are input will be described.

First, when the odd field K2 is input, the motion compensation prediction signal 1 from the odd field K1 of the previous frame stored in the odd field memory 8 is used.
04a is input to the selector 11.

Similarly, the even field G1 of the previous frame stored in the even field memory 9 is input to the selector 11 as the motion compensation prediction signal 104b. Further, the data of K1 and G1 are input to the interpolation unit 10, and after the interpolation processing shown in FIG.
This is input to the selector 11 as the motion compensation prediction signal 104c. Inside the selector 11, these three types of motion compensation prediction signals 104a, 104b, 104
By inputting c and input image signal 101 at the same time, a prediction signal that minimizes the error signal power is selected by comparing each of them.

Similarly, the motion compensation prediction signal 104a based on the odd field K1 stored in the odd field memory 8 and the even number field stored in the even field memory 9 are also stored in the even field G2 of the currently input frame. Motion compensated prediction signal 1 based on field G1
04b, and a motion compensated prediction signal 104c obtained as a result of inputting the motion compensated prediction signals 104a and 104b based on both of these fields and performing an interpolation process,
The selector 11 selects a prediction signal that minimizes the error signal power of the encoding target block of the input image signal 101.

Embodiment 2 In the first embodiment, the interpolation section is provided, and the interpolation section includes the motion compensation prediction signal 104 from the odd field memory 8 and the even field memory 9.
Although the case where the interpolation processing is performed based on a and 104b and the motion compensated prediction signal 104c is output has been described, a case where the interpolation unit 10 does not exist as shown in FIG. 6 may be used. In this case, the selector 11 stores the previous odd field K1 stored in the odd field memory and the previous even field G1 stored in the even field memory 9.
The selector 11 generates the motion compensated prediction signal by using the two types of motion compensated prediction signals 104a and 104.
The prediction signal that minimizes the error signal power is selected from b.

Embodiment 3 FIG. Further, in the first embodiment, a simple averaging is used as the interpolation unit. However, by using a weighted averaging unit considering the field distance as shown in FIG. Can be realized. FIG. 7 is an example of a block diagram of the interpolation unit 10. The motion compensation prediction signal 104a from the odd field is multiplied by the weight α based on the distance from the field to be coded, and the motion compensation prediction signal 104b from the even field is weighted based on the distance from the field to be coded. Multiply by β and take the average,
This output is used as the interpolation prediction signal 104c.

Next, specific weight values of the interpolation unit 10 according to the third embodiment will be described with reference to FIG. As shown in FIG. 5, assuming that T is a unit time for inputting an odd field or an even field, there is a time difference of 2T between the odd field K1 and the odd field K2. There is a time difference of T between the even field G1 and the odd field K2. Therefore, it is possible to determine the weights α and β using this time difference. For example, since the odd field K1 has a time distance of 2T, α is set to 1 and the even field G
1 has a time distance between the odd fields K2 and T,
By setting β to 2, it becomes possible to increase the weight for a thing with a shorter time distance. Similarly, since the odd field K1 has a time distance of 3T with respect to the even field G2 and the even field G1 has a time distance of 2T, when weighting the even field G2, α is set to 2 and β is set to 3 By doing so, it becomes possible to assign a weight proportional to the time distance.

Embodiment 4 In the third embodiment, the case where the weights α and β are determined based on the time distance in the interpolation unit has been described. However, regardless of the time distance, for example, the weight α applied to the odd field is always applied to the even field β. Weighting such as making the weight larger or smaller may be performed. Also,
In the third embodiment, the case where the weights α and β used for the odd field and the weights α and β used for the even field are different is shown. However, the weights used for the odd field and the even field may be equal. Further, in the third embodiment, the case where the weights α and β are simply used has been described. However, the weights may be changed by other coefficients, for example, a coefficient having a quadratic function or a function having a special characteristic. It does not matter if you decide. Further, the weights α and β are not limited to the case where only one type is held, and a plurality of types of weights α and β are prepared according to the type of input signal or the characteristics of the input signal, and are switched. You can use it.

Embodiment 5 Next, another embodiment of the present invention will be described with reference to FIG. In FIG. 8, reference numeral 12 denotes a blocking selection unit for selecting whether to block the prediction error signal individually for odd and even fields or to block including pixels of both odd and even fields, and 13 for the blocking selection unit. Is a blocking decomposer that decomposes the block into the original field according to the output of the blocking selector, and 400 is a blocking part. This is the same as that shown in FIG.

FIG. 9 is an example of a block diagram of the blocking selection unit 12. The prediction error signal 105 is stored in the odd field memory 31 if it is an odd field, and is stored in the even field memory 32 if it is an even field.
As shown in FIG. 10 (a) or (b), p = 16,
Considering a block in which q = 16, the individual field blocking unit 33 performs blocking including any pixel of the odd field and the even field in a (p pixel × q line) block, and the coding unit 35 performs coding. I do. Also,
As shown in FIG. 10 (c), the field combining blocking unit 34 blocks both the pixels of the odd and even fields alternately and includes the blocks (p pixels × q lines) included in the block. To encode.
The information amount comparing unit 37 compares the amount of generated information of the coded data coded by the coding unit 35 and the coding unit 36, and outputs a block having a small amount of generated information as the blocking selection signal 111.

FIG. 11 is an example of a block diagram of the blocking component. The prediction error signal 105 is stored in the odd field memory 41 if it is an odd field, and is stored in the even field memory 42 if it is an even field. In accordance with the blocking selection signal 111 output from the blocking selection unit 12, the prediction error signal stored in the odd field memory 41 and the even field memory 42 is converted into a pixel in one of the odd field and the even field (p pixels × q lines). The blocking composing unit 43 switches between performing blocking including the inside of the block of the above or performing blocking including both the pixels of the odd and even fields within the block of (p pixels × q lines),
Output the prediction error signal that is blocked.

FIG. 12 is an example of a block diagram of the blocking decomposition section. The data decoded by the local decoding unit 5 is input to the blocking decomposer 44 and the blocking selection signal 11 output from the blocking selection unit 12.
The decomposition of the blocking is performed according to 1 and stored in the respective field memories 45 and 46. The stored data is output as a decoding error signal 107.

Next, the operation will be described. The prediction error signal 105 obtained by subtracting the prediction signal 110 from the input signal 101 by the differentiator 3 is input to the blocking configuration unit 13 shown in FIG. 11 and the blocking selection unit 12 shown in FIG. The blocking selection unit 12 performs blocking by including only one of the pixels in the odd field and the even field in the block of (p pixels × q lines), or sets both the pixels in the odd and even fields to (p pixels × q lines). A blocking selection signal 111 for selecting whether to perform blocking included in the block of (q line) is output. In the blocking configuration section 13, the blocking selection signal 1
In accordance with No. 11, either the field-specific blocking or the block including pixels of both fields is defined as (p ×
q) The processing is performed in units of blocks, and the block-divided signal is input to the encoding unit 4. The encoding unit 106 performs orthogonal transform and performs a scalar-quantized transformation on the encoded data 106.
Are sent to the local decoding unit 5 and the multiplexing unit 8.

In the local decoding unit 5, after inverse scalar quantization and inverse orthogonal transformation are performed, the blocking is decomposed into fields according to the blocking selection signal 111. The blocking decomposing unit 14 shown in FIG. Decomposition results in a decoded difference signal 107. The prediction signal 110 is added to the decoded difference signal 107 by the adder 6, and the obtained local decoded signal 108 is stored in the first field memory 8 if it is an odd field, and is stored in the second field memory 9 if it is an even field. It is held and used to detect the motion of each field of the next frame.

In the above embodiment, p = 1
6, q = 16 has been described as one block unit, but the values of p and q are as follows as the block size n × m used by the motion detection unit 2 described in the first embodiment. It is desirable to have a relationship. p = n, q = 2m Since DCT transform is usually performed with 8 pixels × 8 lines, four of these are combined into 16 pixels × 16 lines as a block size in the blocking configuration unit. (Ie, as the values of p and q). As in this example, p = n means that n =
With 16 pixels and q = 2m, m = 8. That is, in order for the motion detection unit 2 to perform motion detection on each of the odd field and the even field, it is desirable that the number of lines is reduced to half, that is, eight. On the other hand, in the blocking configuration unit, since blocking such as combining the odd field and the even field is conceivable, it is desirable to configure 16 lines as one block as the number required when both are configured. .

Embodiment 6 FIG. In the fifth embodiment, the blocking selection is performed by comparing the amount of information generated due to the difference in blocking as shown in FIG. 9 in the selection of the blocking. However, the coding as shown in FIG. By selecting the blocking by comparing the qualities, it is possible to perform the encoding based on the encoding quality.

FIG. 13 is an example of a block diagram showing the configuration of the blocking selection unit 12. As shown in FIG. Prediction error signal 105
Is an odd field memory 51 if it is an odd field.
If the field is an even field, the even field memory 52
Is stored in The field individual blocking unit 53 performs block processing including only pixels in either the odd field or the even field in a block of (p pixels × q lines), and performs encoding / decoding in the encoding / decoding unit 55. .
In addition, the field combining blocking unit 54 performs blocking including the pixels of both fields inside the block of (p pixels × q lines), and the encoding / decoding unit 56 performs encoding / decoding.
Perform decryption. The difference between the encoded / decoded data when performing individual field blocking and the data immediately before encoding, and the encoded / decoded data and the data immediately before encoding when performing field synthesis blocking Is compared with the difference by the error comparator 59 to select a block having a small error, and to select the blocking selection signal 11
Output as 1.

Embodiment 7 In the fifth embodiment, the comparison of the amount of generated information is performed for selecting a block.
Although the blocking selection is performed by comparing the coding errors, the coding using orthogonal transform or the like is performed by selecting the blocking by comparing the content frequency components due to the difference of the blocking as shown in FIG. 14 described below. It is possible to perform more efficient encoding when performing the encoding.

FIG. 14 is an example of a block diagram showing the configuration of the blocking selection unit 12. As shown in FIG. Prediction error signal 105
Is an odd field memory 61 if it is an odd field
If the field is an even field, the even field memory 62
Is stored in The field individual blocking unit 63 performs block formation including only pixels in the odd field and the even field in a block of (p pixels × q lines), and performs frequency analysis in the frequency analysis unit 65 as shown in FIG. 15, for example. Do. The field synthesis blocking unit 64 performs block formation including the pixels of both fields (p pixels × q lines) inside the block, and frequency analysis is performed by, for example, a frequency analysis unit 66 as shown in FIG. A block having less high-frequency components is selected when the field is divided into individual blocks and when the field synthesis is blocked, and is output as the blocking selection signal 111.

FIG. 15 is an example of a block diagram showing the configuration of the frequency analyzers 65 and 66. A signal obtained by individually blocking the odd and even fields from the field individual blocking unit 63 or a signal obtained by blocking the pixels including both the odd and even fields from the field configuration blocking unit 64 is input. The converter 68 converts the signal in the pixel area into a signal in the frequency domain. High-frequency components are extracted from the converted frequency-domain signal by a high-frequency component selector 69, and the high-frequency component adder 70 accumulates the extracted high-frequency components. The accumulated high-frequency components are compared in a high-frequency component comparison section 67, and a block having a small high-frequency component is selected.

FIG. 16 shows an example of components accumulated by the high-frequency component adder in the frequency domain signal subjected to the orthogonal transformation. Here, as an example, eight components having the highest frequency component of the vertical frequency component are selected.

Embodiment 8 FIG. In the above embodiment, although the selection information of the prediction signal and the selection information of the blocking are not used as the encoding unit 4, the output of the selector 11 which is the selection signal of the prediction signal as shown in FIG. Is input to the encoding unit 4 and the encoding characteristics are controlled using the selected prediction signal and the selected blocking information, thereby enabling more fine-grained control. High quality coding can be realized.

As described above, in the first embodiment, the input image signal obtained by the interlaced scanning is predictively coded by using the motion compensation. A motion detecting means for obtaining a displacement amount for performing motion compensation prediction in block units (n and m are positive numbers) individually (n pixels × m lines) from both the odd and even fields of the already coded frame; , A first prediction signal 104a obtained by motion compensation from the odd field, and a second prediction signal 104 obtained by motion compensation from the even field.
b and a plurality of candidates including a third prediction signal 104c obtained by interpolating the first and second prediction signals, a prediction signal which gives an optimum prediction is selected by the selector 11 and The prediction error signal output means for obtaining the difference between the input signal and the field and outputting it as a prediction error signal has been described.

In the first embodiment, the interpolating means for obtaining the third predicted signal is a means for simply averaging the first predicted signal and the second predicted signal. The field / frame coding scheme has been described.

As described above, by simply adding and averaging both the odd-numbered and even-numbered motion-compensated fields to generate an interpolation signal of the prediction signal, the scale of the hardware is minimized and the prediction efficiency is reduced. Good encoding can be achieved.

In the third embodiment, the interpolating means for obtaining the third prediction signal converts the first prediction signal and the second prediction signal into a field used for prediction and a field to be encoded. The adaptive field / frame coding method characterized in that it is a weighted averaging means taking into account the temporal distance.

As described above, both the odd-numbered and even-numbered motion-compensated fields are interpolated by weighted averaging in consideration of the temporal distance between the field used for prediction and the field to be encoded. By generating, encoding with very good prediction efficiency can be realized.

In the fifth embodiment, in order to encode the prediction error signals for the odd and even fields of the input image signal in block units (p pixels × q lines) (p and q are positive numbers), Blocking is performed by including only one of the pixels of the odd and even fields in the (p pixel × q line) block, or the pixels of both the odd and even fields are stored in the (p pixel × q line) block. An adaptive field / frame coding method characterized by comprising means for coding while switching whether or not to perform blocking including the adaptive field / frame coding method.

Further, in the fifth embodiment, the block-forming means for coding while switching the above-mentioned blocks only uses one of the pixels of the odd and even fields (p pixels × q
The generation information of the coding is obtained by the case where the block is included inside the block of (line) and the case where the block is formed by including both the pixels of the odd field and the even field within the block of (p pixel × q line). An adaptive field / frame coding scheme characterized by having a selection means for selecting a small amount of blocking has been described.

Further, in the sixth embodiment, the block-forming means for coding while switching the above-mentioned blocks only uses one of the pixels of the odd and even fields (p pixels × q
The coding error is small between the case where the block is included inside the block of (line) and the case where both the pixels of the odd and even fields are included within the block of (p pixel × q line). 2. The adaptive field / frame coding method according to claim 1, further comprising means for selecting blocking.

Further, in the seventh embodiment, the blocking means for coding while switching the above-mentioned blocks only uses one of the pixels in the odd and even fields (p pixels × q
The signal to be encoded is divided into a case where the block is included inside the block of (line) and a case where both the pixels of the odd and even fields are included in the block of (p pixel × q line). The adaptive field / frame coding method, which has a selection unit for selecting a block containing a small amount of high-frequency components included in the above, has been described.

In the eighth embodiment, (p pixels ×
When coding including quantization of an orthogonal transformer and a transform coefficient is used in a block-by-block encoding unit of (q line), the quantization characteristic of the transform coefficient is determined according to the selected prediction signal and the selected blocking. The adaptive field / frame coding method characterized in that coding is performed while switching is described.

Embodiment 9 FIG. Note that, in the above embodiment, the case where the input image signal 101 is configured by a frame having an odd field and an even field has been described. However, the odd field and the even field are examples, and the odd or even field is named. I don't care. The present invention can be similarly used in the case where one frame is divided into a plurality of fields, and the odd field and the even field are examples. For example, not only odd and even numbers, the first and second lines are set as the first field, and the third and fourth lines are set as the second field.
The present invention can also be applied to a case where the fifth and sixth lines are stored as two fields, such as the first field and the seventh and eighth lines are stored as two fields. It is. Also, an odd field, an even field, or a first field,
The second field is not limited to the case where the field is divided into two types of fields, and may be the case where the field is divided into two or more types, that is, three or four types of fields. In this case, a field memory corresponding to each field is provided, and the above-described processing is performed on each field.

Embodiment 10 FIG. Also, in the above embodiment, the blocking selection unit has shown two cases of blocking only one of the pixels of the odd and even fields and blocking the pixels of both the odd and even fields. However, when the field is not limited to an odd number or an even number, and there are two or more types, it is possible to block by various combinations, and (a), (b), and (c) shown in FIG. The block in the parentheses) is an example, and it goes without saying that there are various block configuration methods when a block is configured.

Embodiment 11 FIG. Further, in the above embodiment, as shown in FIG. 8, the blocking means is present together with the prediction error signal output means and the motion detection means, but the blocking means 400
The third and fourth aspects of the invention are effective even when the other parts are as described in the conventional example.

[0056]

As described above, according to the present invention,
The already decoded image signal is divided into a plurality of fields and stored, and a plurality of prediction signals for predicting a change in the image signal based on the field signal are output, and an interpolation prediction signal is obtained from at least two different prediction signals. Generating, selecting one prediction signal from the plurality of prediction signals and the interpolation prediction signal, and generating and outputting a decoded image signal based on the decoded prediction error signal and the selected prediction signal. The decoding can be correctly performed even when the motion compensation prediction coding is performed on the side using the signal obtained by interpolating the prediction signals from both fields of the already coded frame as the prediction signal. As a result, on the encoding side, motion is individually searched for from each field of the already encoded frame in order to predict each field, and adaptation from the plurality of types of retrieved motion compensated prediction signals (and their interpolation signals) is performed. By performing prediction, an efficient and stable encoded image can be obtained. On the decoding side, an efficient and stable encoded image that has been adaptively predicted and encoded on the encoding side in such a manner is correctly decoded. It becomes possible.

[Brief description of the drawings]

FIG. 1 shows an adaptive field / according to one embodiment of the present invention.
FIG. 3 is a block diagram illustrating a configuration of a frame encoding scheme.

FIG. 2 is a diagram illustrating an embodiment of an input image signal.

FIG. 3 is a block diagram illustrating an example of a configuration of an interpolation unit according to the embodiment of the present invention.

FIG. 4 is a diagram for explaining the operation of a motion detection unit.

FIG. 5 is a diagram for explaining an operation using a motion compensation prediction signal according to one embodiment of the present invention;

FIG. 6 is a block diagram showing a configuration of an adaptive field / frame coding scheme according to another embodiment of the present invention.

FIG. 7 is a block diagram illustrating another example of the configuration of the interpolation unit.

FIG. 8 is a block diagram showing a configuration of an adaptive field / frame coding scheme according to another embodiment of the present invention.

FIG. 9 is a block diagram illustrating an example of a configuration of a blocking selection unit.

FIG. 10 is a diagram illustrating a configuration example of a block that is selected as a candidate by a blocking selection unit;

FIG. 11 is a block diagram illustrating an example of a configuration of a blocking configuration unit.

FIG. 12 is a block diagram illustrating an example of a configuration of a blocking decomposition section.

FIG. 13 is a block diagram illustrating another example of the configuration of the blocking selection unit.

FIG. 14 is a block diagram illustrating another example of the configuration of the blocking selection unit.

FIG. 15 is a block diagram illustrating an example of a configuration of a frequency analysis unit.

FIG. 16 is a diagram illustrating an example of accumulated frequency components.

FIG. 17 is a block diagram showing an example of another configuration of the present invention.

FIG. 18 is a block diagram illustrating a configuration of a conventional encoding method.

[Explanation of symbols]

Reference Signs List 1 frame memory 2 motion detection unit 3 differentiator (same for 57 and 58) 4 encoding unit 5 local decoding unit 6 adder 7 multiplexing unit 8 first field memory 9 second field memory 10 interpolation unit 11 selector 12 Blocking selection unit 13 Blocking configuration unit 14 Blocking decomposition unit 31 Odd field memory (41, 45, 51, 6)
32) Even field memory (42, 46, 52, 6)
33) Individual field blocking section (53 and 63 also) 34 Field combining blocking section (54 and 64) 35 Encoding section (36) 37 Information amount comparing section 43 Blocking constructor 44 Blocking decomposer 55 encoder / decoder (same for 56) 59 error comparator 65 frequency analyzer (same for 66) 67 high-frequency component comparator 101 interlaced input image signal 102 neighborhood block 102a neighborhood block in first field 102b Two-field neighboring block 103 Motion vector 104 Motion-compensated predicted signal 104a Motion-compensated predicted signal obtained from the first field 104b Motion-compensated predicted signal obtained from the second field 104c Interpolated motion-compensated predicted signal 105 Prediction error signal 106 Coding data 107 decoding error signal 108 local decoding signal 109 transmission line 110 selected prediction signal 111 blocking selection signal 200 motion detection means 300 prediction error signal output means 400 blocking means 500 encoding means

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-4-252690 (JP, A) JP-A-4-288790 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) H04N 7/32

Claims (4)

(57) [Claims] 1. An input image signal comprising odd / even fields interlaced and obtained from a plurality of prediction signals.
The encoded data of the prediction error signal is a motion compensated prediction is generated by coding a difference between the prediction signal which is to decode based on the motion vectors used in motion compensated prediction, in a local video decoder there are, erroneous decoded prediction by decoding the encoded data of the prediction error signal
Decoding means for generating a difference signal ; a decoded prediction error signal decoded by the decoding means;
Adding means for adding the prediction signal and outputting a decoded image signal; and outputting the decoded image signal output by the adding means as an odd number /
From a field memory that divides and stores even fields, and a decoded image signal stored in the field memory,
Fi odd / even field based on the motion vector
A prediction signal output means for outputting a Rudo prediction signal, field prediction signal of odd / even field output from the field memory by the prediction signal output means
And interpolation means for generating an interpolated predictive signal to input No., field prediction signal of odd / even field output from the field memory by the prediction signal output means
No. and output hand the predicted signal among the interpolated predictive signal outputted from said interpolating means, and outputs it to the adding means
A local image decoder comprising: a stage ; 2. The local image decoder according to claim 1, wherein said interpolation means generates an interpolation prediction signal by calculating an average of prediction signals of odd / even fields. 3. An input image signal comprising odd / even fields interlaced and obtained from a plurality of prediction signals.
The encoded data of the prediction error signal is a motion compensated prediction is generated by coding a difference between the prediction signal which is to decode based on the motion vectors used in motion compensated prediction, in a local image decoding method there are, erroneous decoded prediction by decoding the encoded data of the prediction error signal
To generate a difference signal, the decoded prediction error signal and the
The decoded image signal is output by adding the prediction signal to the decoded image signal. The decoded image signal is divided into odd / even fields and stored. The stored decoded image signal is divided into odd / even fields based on the motion vector. Feel
Outputs de prediction signal, and generates an interpolated predictive signal by entering the field prediction signal of the odd / even field, the predicted signal among the field prediction signal and the interpolated predictive signal of the odd / even field , and outputs it to the adding means, a local image decoding method, characterized in that. 4. The local image decoding method according to claim 3, wherein an interpolation prediction signal is generated by calculating an average of prediction signals of odd / even fields.