JP4380498B2 - Block distortion reduction device - Google Patents

Description

  The present invention relates to a block distortion reduction apparatus, and more particularly to decoding of encoded image data obtained by compressing and encoding an image by a block unit encoding method, and generating block distortion (block) generated at a block boundary of the decoded image data. The present invention relates to a block distortion reducing device that reduces noise.

  In image compression coding such as the MPEG (Moving Picture Experts Group) standard, image data is divided into block units of a predetermined size composed of a plurality of predetermined pixels adjacent in the horizontal and vertical directions, and within the divided blocks. A method (block unit coding method) in which compression coding is performed by performing processing such as discrete cosine transform (DCT) in units of blocks using the high correlation between adjacent pixels of Yes.

  In the case of image data encoded by this block unit encoding method, when the compression rate is increased in order to reduce the amount of information, the boundary between adjacent blocks due to quantization errors is added to the decoded image data. Block distortion (block noise), in which (block boundary) appears as distortion (tone difference), is likely to occur. Therefore, in order to reduce the block distortion, a block distortion reduction apparatus that applies a noise reduction filter to a block boundary where block distortion occurs has been proposed (see, for example, Patent Document 1).

  FIG. 8 shows a block diagram of an example of a conventional block distortion reducing apparatus described in Patent Document 1. In the same figure, the input video signal is supplied to the isolated differential point extraction circuit 1 where it is differentiated into a differential signal, and then the differential value of the adjacent pixel is compared and only the differential value protruding from the adjacent pixel is obtained. Extracted as an isolated differential point. The isolated differential data output from the isolated differential point extraction circuit 1 is supplied to the block noise detection circuit 2, where block noise (block distortion) is utilized by utilizing the fact that the isolated differential points have high correlation in the horizontal and vertical directions. ) Is detected and supplied to the switching terminal of the switch 4.

  When the block noise detection signal indicating that there is a lot of block noise is supplied from the block noise detection circuit 2, the switch 4 selectively outputs the isolated differential data output from the isolated differential point extraction circuit 1 to the filter 5, and in units of frames. When there is little block noise and no block noise detection signal is input, a 0 level signal is selectively output to the filter 5. The filter 5 outputs a correction signal for correcting the signal level difference at the boundary between the rectangular block in which block noise is generated and the adjacent block.

On the other hand, the input video signal is delayed by a predetermined time delay circuit 3 for adjusting the timing of the output correction signal of the filter 5 and then supplied to the adder 6 where it is added to the output correction signal of the filter 5. As a result, when block noise is generated from the adder 6, the gradation difference at the block boundary between adjacent blocks is smoothed and the block noise is hardly noticeable and is output to the output terminal 7. Is done. This avoids filtering at positions other than block noise and suppresses adverse effects. For frames with a lot of block noise, isolated differential data is selectively output from the switch 4 and a frame with little block noise. Because a 0-level signal is output from the switch 4, block noise reduction processing is performed for frames with a lot of block noise, and block noise reduction processing is turned off for frames with a low block noise. Thus, image quality deterioration associated with block noise reduction processing is suppressed.

JP 2001-119695 A (FIG. 5)

  However, the conventional block distortion reduction device described in Patent Document 1 described above is erroneously determined as block noise in an image contour portion or the like in an image with little block noise (block distortion), and performs unnecessary filtering. However, since only the on / off of the block noise reduction processing is controlled, there is a problem that visual unnaturalness is given at the time of switching.

  The present invention has been made in view of the above points, and an object of the present invention is to provide a block distortion reduction device that can suppress erroneous determination and reduce visual unnaturalness in an image with little block distortion.

In order to achieve the above object, the present invention decodes block-unit encoded image data obtained by compressing and encoding image data in units of blocks composed of a plurality of predetermined pixels adjacent in the horizontal and vertical directions. In the block distortion reduction device for reducing block distortion generated at the boundary between adjacent blocks with respect to the decoded image data, the pixel values around the block distortion of the decoded image data are smoothed. performs, p · q pieces (p, q are each 2 or greater natural number) filter, decoded adjacent block boundaries of the entire in one image of the image data having the strength properties of the different smoothing mutually in a first block distortion amount detecting means for performing detecting the amount of the block distortion generated in one image unit, the decoded image data of the adjacent block A second block distortion amount detecting means for performing and detecting the distortion of the block generated between pixels adjacent the field in pixel units of a block of first image units detected by the first block distortion amount detecting means A calculation unit that calculates an average value or an intermediate value of distortion amounts for a plurality of consecutive images, and a decoded image data of one image that is input next to the plurality of consecutive images calculated by the calculation unit. In order to perform switching for supplying the input image data to one filter selected from the p · q filters in units of pixels, the p · q filters are the filters having the maximum smoothing intensity characteristics. is p pieces each q sets of filter according to the strength of the strength properties until the minimum filter from the average value or the intermediate value calculated by the calculating means large becomes large indeed strength properties filter Select Ranaru group, selected according to the amount of block distortion of the detected image-containing units by a single filter, the second block distortion amount detecting means of the selected set of p pieces of filter And switching means for performing switching, and outputting image data with reduced block distortion from a selected filter.

According to the present invention, when p · q filters are used as q sets of filters in accordance with the strength of the intensity characteristic from the filter having the maximum smoothing intensity characteristic to the filter having the smallest smoothing intensity characteristic, A group consisting of filters having greater intensity characteristics as the average value or intermediate value of the distortion amount increases, and one of the p filters in the selected group is selected as the second block distortion amount detection means. Since the selection is made according to the amount of block distortion of one pixel unit detected by , block distortion such as an uncompressed image does not affect the distortion reduction effect in an image with much block distortion. Even in the case of an image that does not appear at all, and there is a pattern (such as an outline) that may be erroneously detected as block distortion, this can be prevented from being unnecessarily filtered.

According to the present invention, in an image with a large amount of block distortion, the effect of reducing the distortion is not affected, and in an image with a small block distortion, erroneous determination is suppressed and the amount of block distortion greatly changes for each image. Even in this case, visual unnaturalness due to frequent filter switching can be reduced.

  Next, each embodiment of the present invention will be described with reference to the drawings.

(First embodiment)
FIG. 1 shows a block diagram of a first embodiment of a block distortion reducing apparatus according to the present invention. In the figure, a decoder 11 decodes block-unit encoded image data input via an input terminal 10. The block distortion amount detection unit 13 in units of frames and the block distortion detection unit 12 in units of pixels use the same block distortion detection method as the conventional method for the decoded image data, Each block distortion amount is detected .

The filter switch 14 includes a block distortion amount detection signal detected in units of frames by the block distortion amount detection unit 13 in units of frames, and a block distortion detection signal detected in units of pixels by the block distortion detection unit 12 in units of pixels. Is selected as a selection control signal, a selection operation is performed with priority given to a block distortion amount detection signal detected in frame units, which will be described later, and the decoded image data input to the common terminal is converted into n filters 15 ₁ to 15 ₁ . 15 Select and output to one filter selected and designated among _n .

The filters 15 ₁ to 15 _n are edge-preserving smoothing filters for smoothing the pixel values around the block distortion, but the smoothing strength characteristics are different from each other. For example, the filter 15 ₁ is strongest strength properties of smoothing of the pixels around the block distortion location, number of pixels is most often the value is changed, the following, the filter 15 _2, 15 _3, ... order of strength of in smoothing gradually weakened, the strength characteristics of the filter 15 _n is smoothed weakest, there is a characteristic value only left and right pixels around the block distortion position is changed. Therefore, the values of pixels in a wide range of positions are changed so that the filter having stronger smoothing intensity characteristics becomes smoother with respect to the pixels around the block distortion.

Here, when a general low-pass filter (LPF) is used as the above-described filters 15 ₁ to 15 _n , there arises a problem that the resolution is lowered when the edges are rounded. Therefore, as described above, edge-preserving smoothing filters are used as the filters 15 ₁ to 15 _n in the present embodiment. This edge-preserving smoothing filter includes an ε filter, a ripple filter, etc. According to the literature (Takahiko Fukiuki, “Suppression of DCT Quantization Distortion by Ripple Correction Filter”, 2002, Winter Conference of the Institute of Image Information and Television Engineers) Are known.

  Next, the operation of the present embodiment will be described. The block-unit encoded image data input via the input terminal 10 is supplied to the decoder 11 and decoded. The decoded image data obtained by decoding by the decoder 11 is supplied to the block distortion amount detection unit 13 in units of frames and the block distortion detection unit 12 in units of pixels, respectively, while being connected to a common terminal of the filter switch 14. Supplied.

  The block distortion amount detection unit 13 in units of frames detects the block distortion amount of the input decoded image data in units of frames, and supplies the block distortion amount detection signal to the filter switch 14 as a selection control signal. On the other hand, the block distortion detection unit 12 in units of pixels detects block distortion of the input decoded image data in units of pixels and supplies the block distortion detection signal to the filter switch 14 as a selection control signal.

The filter switch 14 determines a rough filter selection range for selecting which filter from the filters 15 ₁ to 15 _{n according} to the block distortion amount in the frame unit indicated by the block distortion amount detection signal in the frame unit. For example, as this filter selection range, when n = 9, the set of filters 15 ₁ to 15 ₃ , the set of filters 15 ₄ to 15 ₆ , and the set of filters 15 ₇ to 15 ₉ are respectively set as rough filter selection ranges, One set (filter selection range) is determined (selected) from the block distortion amount in units of frames.

  At this time, if the amount of block distortion per frame is large, select a set of filters with strong smoothing intensity characteristics. If the amount of block distortion is small, select a set of filters with low smoothing intensity characteristics. When the amount of block distortion is medium, a filter group having a medium smoothing intensity characteristic is selected.

  Subsequently, the filter switch 14 selects one of a plurality of filters of one set (filter selection range) previously determined according to the block distortion amount in pixel units indicated by the block distortion detection signal in pixel units. Select filters. Accordingly, when the amount of block distortion in units of frames is small, the filter switch 14 selects from a set of filters having a weak smoothing intensity characteristic of the filter to be used no matter how large the block distortion in units of pixels is detected. It will be.

In this way, the filter switch 14 supplies the decoded image data from the decoder 11 to one selected filter among the filters 15 ₁ to 15 _n . One of the filters 15 ₁ to 15 _n to which the decoded image data is supplied from the filter switch 14 performs an edge preserving smoothing process and outputs image data with reduced block distortion as an output terminal 16 is output.

Here, as the above 15 ₁ to 15 _n , for example, the edge preserving smoothing process when the ε filter which is a nonlinear filter is used will be described with reference to the flowchart of FIG. First, the pixel to be processed is A, the adjacent pixel of the pixel A is a (x), and these values are taken in (step S1), and the difference value diff (= a (x) −A) is calculated (step S1). S2). Subsequently, it is compared whether or not the difference value diff is larger than a predetermined first threshold value THR (step S3). When diff> THR, the difference value diff is set to the first threshold value THR which is an upper limit value. (Step S4).

  On the other hand, when a comparison result of diff ≦ THR is obtained in step S3, it is compared whether or not the difference value diff is smaller than the second threshold value −THR (step S5), and if diff <−THR, the difference is calculated. The value diff is set to the second threshold value -THR which is the lower limit value (step S6). Further, when the comparison result of diff ≧ −THR is obtained in step S5, that is, when the difference value diff is between THR and −THR, or when the upper limit value is set in step S4, or in step S6. When the lower limit value is set, the difference value diff at that time is substituted into the variable sum_diff (step S7), and the processing of steps S1 to S7 is repeated for all the target adjacent pixels (step S8).

  In this way, a value after_filter is calculated by adding the value A of the target pixel to the average value obtained by dividing the variable sum_diff obtained for all of the target adjacent pixels by the number of adjacent pixels n (step S9). Is output as a filter output.

  Thereby, for example, among the pixels a1 to a5 arranged in the horizontal direction shown in FIG. 5, a 5-tap ε filter whose processing target is the central pixel a3 is obtained by the following equation.

上記のεフィルタは図６のように、入力ｉｎに対し、出力ｏｕｔが非線形となり、上式における画素差分絶対値ＡＢＳ（ａｉ−ａ３）が閾値ＴＨＲより小さい場合に、平滑化の対象となる。すなわち、閾値ＴＨＲによりフィルタ強度を制御できる。よって、ブロック歪み量検出部１３からの歪み量情報により、歪み量が大きい場合にはエッジ保存型平滑化フィルタ強度を強く、適用範囲を広くするようにフィルタの切り替えを行うことで、切り替え時の不自然さの少ないブロック歪みの軽減を行うことができる。

As shown in FIG. 6, the ε filter is a target to be smoothed when the output out becomes nonlinear with respect to the input in and the pixel difference absolute value ABS (ai−a3) in the above equation is smaller than the threshold value THR. That is, the filter strength can be controlled by the threshold value THR. Therefore, according to the distortion amount information from the block distortion amount detection unit 13, when the distortion amount is large, the edge preserving smoothing filter strength is increased, and the filter is switched so as to widen the application range. Block distortion with less unnaturalness can be reduced.

Thus, according to the present embodiment, not only on / off of one filter as in the conventional block distortion reducing device described in Patent Document 1, but also by switching the filters 15 ₁ to 15 _n having several levels of strength. Smooth filtering with reduced visual effects is possible.

  In addition, according to the present embodiment, the block distortion amount in units of frames is detected, and the block distortion amount is used as main determination information for filter selection. Compared to using a pixel change value to select a filter based on the presence and intensity of block distortion at each block boundary, partially unblocked images such as uncompressed images. Even if there is a pattern (such as an outline) that is erroneously detected as distortion, it is possible to prevent unnecessary filtering.

(Second Embodiment)
Next, a second embodiment of the present invention will be described. FIG. 2 shows a block diagram of a second embodiment of a block distortion reducing apparatus according to the present invention. In the figure, the same components as those in FIG. In the second embodiment shown in FIG. 2, a block boundary distortion strength detection unit 21 that detects a block distortion strength using a pixel change value in the vicinity of the block boundary is added to the configuration of the first embodiment. Thus, the information is also characterized in that the filter is switched in addition to the judgment material.

For example, when the filters 15 ₁ to 15 _n are the ε filters, the filter strength is strongly applied to block boundaries having a large distortion strength. As a result, filter switching in units of block boundaries can be realized, so that frame-unit information from the block distortion amount detection unit 13 in units of frames is used for various types of information such as being used for frame switching of the filters 15 ₁ to 15 _n . More optimal selection application can be performed.

  The block boundary distortion strength detection unit 21 performs block distortion strength detection on the input decoded image data using the block boundary and neighboring pixel difference values for each block boundary, and outputs the result to the filter switch 22. To do.

  FIG. 7 is a diagram showing the vicinity of a unit block boundary in block coding. When the block boundary distortion strength detection unit 21 performs block distortion strength detection for the block boundary 81, detection is performed using a 1, a 2, a 3, i 1, i 2, i 3 that are adjacent pixels of the block boundary 81.

The pixel difference absolute value between pixels (a1-i1) sandwiching the block boundary 81 is represented by bound0, and the pixel difference absolute value between adjacent pixels of the block boundary neighboring pixels is represented by diff2 to diff5. When these difference absolute values are large, it indicates that the pixel value changes drastically between the pixels. Therefore, the filter switch 22 is the average of the boundary pixel difference absolute value bound0 and the neighboring pixel difference absolute value.
AveDiff_a = (diff2 + diff3 + 1) / 2
AveDiff_i = (diff4 + diff5 + 1) / 2
Compare When bound0 is large with respect to AveDiff_a and AveDiff_i, it indicates that the block boundary 81 has undergone a large change compared to its vicinity. Since the block distortion is a noticeable distortion because the block boundary 81 is greatly changed compared to the adjacent portion, the block distortion can be detected by detecting an isolated pixel value change between the block boundary pixels. it can.

  The filter switch 22 determines the distortion strength based on the magnitude of bound0 and the difference from AveDiff_a and AveDiff_i, and selects a filter corresponding to the determined distortion strength. However, the filter switch 22 determines a set of filters that most preferentially selects the block distortion amount in the frame unit detected by the block distortion amount detection unit 13 in the frame unit, and the distortion intensity detection unit 21 at the block boundary. The detection result of the block boundary distortion intensity detected in (1) determines the filter to be selected in the lowest priority order.

As a result, in the present embodiment as well as in the first embodiment, not only the on / off of one filter as in the conventional block distortion reducing device described in Patent Document 1, but also the filter 15 having several levels of strength. By switching from ₁ to 15 _n , smooth filtering with suppressed visual influence becomes possible.

(Third embodiment)
Next, a third embodiment of the block distortion reducing apparatus according to the present invention will be described. FIG. 3 shows a block diagram of a third embodiment of a block distortion reducing apparatus according to the present invention. In the figure, the same components as those in FIG.

In the third embodiment shown in FIG. 3, information on the amount of distortion obtained by the block distortion amount detection unit 13 in units of frames is sequentially transmitted at the timing of the frame pulse 33 generated at every frame switching. Are supplied to _m holding units 32 _{1 to} 32 _m to thereby hold a block distortion amount for m frames (m is an arbitrary natural number), respectively, and frames for m frames from the holding units 32 _{1 to} 32 _m. The block distortion amount in units is supplied to the average value calculation unit 34 in parallel to calculate the average value, and the average distortion amount is supplied to the filter switch 31.

The filter switch 31 converts the image data of the (m + 1) th frame into the filters 15 ₁ to 15 based on the average value of the block distortion amount detected by the average value calculation unit 34 in the frame unit from the first frame to the mth frame. Among the set of filters corresponding to the average value among 15 _n , one filter is selected and supplied according to the detected amount of block distortion in pixel units. Thus, according to the present embodiment, frequent filter switching can be avoided and visual unnaturalness can be suppressed even when the block distortion amount changes greatly for each frame.

  The present invention is not limited to the above embodiment. For example, the block distortion amount detection in the frame unit in the above embodiment may be performed in the field unit, and in short, in one image unit. Just do it. The average value calculation unit 34 may be replaced with an intermediate value selection unit that selects an intermediate value. It is also possible to select a filter in accordance with the block distortion amount detection result in frame units or field units. Furthermore, the present invention can also be realized by a computer program. In this case, the program can be taken into a computer via a recording medium or a communication line.

It is a block diagram of a 1st embodiment of the present invention. It is a block diagram of the 2nd Embodiment of this invention. It is a block diagram of the 3rd Embodiment of this invention. It is a flowchart of an example for demonstrating the epsilon filter used for this invention. It is a figure of the application pixel of an epsilon filter, and its peripheral pixel. It is a figure which shows the input-output characteristic of an epsilon filter. It is a figure which shows pixel arrangement | positioning around a block boundary. It is a block diagram of an example of a conventional device.

Explanation of symbols

10 Input terminal of image data encoded by block unit encoding 11 Decoder 12 Block distortion detection unit in pixel unit 13 Block distortion detection unit in frame unit 14, 22, 31 Filter switch 15 ₁ to 15 _n filter 16 Block distortion Output terminal for image data with reduced image 21 Distortion intensity detection section at block boundary 32 _{1 to} 32 _m value holding section 33 Frame pulse 34 Average value calculation section

Claims (1)

Decoding block unit encoded image data obtained by compressing and encoding image data in units of blocks composed of a plurality of predetermined pixels adjacent in the horizontal and vertical directions, and for the decoded image data, In a block distortion reduction apparatus that reduces block distortion generated at the boundary between adjacent blocks,
The pixel values around the block distortion of the decoded image data are smoothed, and p · q (p and q are natural numbers of 2 or more) having different smoothing intensity characteristics. Filters,
A first block distortion amount detecting means which performs one image unit detecting the amount of the block distortion generated at the boundary of the adjacent blocks of the whole in one image of the image data which is the decoded,
A second block distortion amount detecting means for performing the pixel units of detecting the distortion amount of the block that occur between the boundaries of adjacent pixels of the adjacent said block of image data said decoding,
Calculating means for calculating an average value or an intermediate value for a plurality of consecutive images of the block distortion amount of one image unit detected by the first block distortion amount detecting means;
The decoded image data of one image input next to the plurality of consecutive images calculated by the calculating means is received as input, and the input image data is selected from the p · q filters in units of pixels. In order to perform switching to be supplied to one filter, the p · q filters are divided into q q pieces according to the strength of the intensity characteristic from the filter having the largest smoothing intensity characteristic to the filter having the smallest smoothing intensity characteristic. When a set of filters is selected, a set consisting of filters having a larger intensity characteristic is selected as the average value or intermediate value calculated by the calculation means is larger, and among the p filters of the selected set, one filter, and a switching means for switching to select depending on the amount of block distortion of the second block distortion amount before being detected by the detecting means Kiga containing units, The block distortion reducing apparatus, wherein the image data with the reduced block distortion is output from the selected filter.

Images