KR20050047871A - Method for fast filtering in moving picture decoding, and apparatus for the same - Google Patents

Abstract

The present invention relates to a fast filtering algorithm for improving image quality in video decoding.

The decoder according to the present invention is characterized by including a filter unit including a temporal filter unit for filtering inter-block boundaries constituting the frame and temporally filtering the filtered frames. On the other hand, it can be implemented including a filter selector to be compatible with the previous video decoder. The decoding method according to the present invention is characterized in that the inter-block boundary constituting the frame is filtered and the filtered frames are temporally filtered.

Description

High speed filtering method in video decoding and apparatus for it {METHOD FOR FAST FILTERING IN MOVING PICTURE DECODING, AND APPARATUS FOR THE SAME}

The present invention relates to a fast filtering algorithm for improving image quality in video decoding.

In recent years, along with digital technology, technology for moving pictures has been rapidly developed.

Basically, the amount of video data is huge and requires a large amount of storage media and a large bandwidth in transmission. For example, a 24-bit true-color image with a resolution of 640 * 480 would require a capacity of 640 * 480 * 24 bits per frame, or about 7.37 Mbits of data. If a video is played at 30 frames per second, it needs 221 Mbits of information per second, and about 1200 Gbits of storage space is needed to store a 90-minute movie. Therefore, a technique for compressing a video is essential and various compressor methods are used. The basic principle of video compression is to reduce the amount of data by eliminating temporal redundancy through frame comparison and spatial redundancy by the conversion process. In most video compression algorithms, when the temporal overlap between frames is removed, motion compensation is performed through motion vectors to achieve high compression ratio.

The compression process is performed by a certain unit such as a macroblock, and the artifacts caused by artificial screen processing appear. Especially when compressed at low bit rates, this unnaturalness is exacerbated. Therefore, in the video decoding process, it is desirable to go through a filtering process to improve the unnaturalness. Filtering can be divided into a loop filter used for encoding and decoding, and a post-processing filter used only for decoding.

In block-based video algorithms such as H.264, loop filters are used, and the filtering process can be largely divided into the blocking phenomenon searching process and the filtering process. In the blocking phenomenon retrieval process, the degree of blocking is expressed differently according to the coding type of each block, and Bs, which is a "Boundary Strength" value, is allocated according to the degree of blocking.

1 illustrates a boundary of a macroblock to be filtered in the H.264 video decoding process. In the drawing, the solid line represents the border of Luma and the dotted line represents the border of Chroma. Bs is assigned to each boundary between neighboring 4 * 4 luma blocks. Different types of filtering are performed according to the boundary strength. In this case, an adaptive filter using coding information such as a quantization parameter is used.

The process of determining the boundary strength will be described with reference to FIG. 2. In order to determine the boundary strength, a boundary between blocks p and q is first input. If block p or q is intra coded and its boundary is the boundary of the macroblock, then Bs is 4. If block p or q is intra coded but its boundary is not the boundary of the macroblock, then Bs is 3.

On the other hand, when both blocks p and q are not intra coded, Bs becomes 2 when a non-zero transform coefficient is included in the block p or q. Otherwise, the prediction of samples of p uses a different number of reference pictures or reference pictures than the prediction of samples of q, or the difference between the x and y components of the motion vector of p and q is greater than 1 pixel. Or Bs equals 1 in all other cases and Bs equals 0 in other cases.

FIG. 3 is a diagram for explaining samples across a horizontal or vertical boundary of a 4 * 4 block. As shown in the figure, a value consisting of four pixels in a horizontally or vertically adjacent 4 * 4 block is p = {p3, p2, p1, p0} and q = {q0, q1, q2, q3}, respectively. The contents of Bs described through 2 can be summarized as follows. That is, for Bs 4, p and q are intra coded, and the boundary between p and q is a macroblock boundary. When Bs is 3, p and q are intra coded, but the boundary between p and q is not the boundary of the macro block. For Bs 2, p and q are inter coded and there are transform coefficients coded in p or q. Bs is 1 when p and q are inter coded, there are no transform coefficients coded in p and q, and the motion vector reference frames of p and q are different. Finally, Bs equals zero does not correspond to the above four.

The filtering process undergoes a different filtering process according to the size of Bs. The filtering process can be classified into four modes according to the difference between the pixels in the block boundary region and the difference between the pixels in the block. Mode 1 is a case where Bs is 0 and does not perform a filtering process.

Mode 2 filters only the pixels p0 and q0 in the block boundary region when Bs is not 0 and the difference between p0 and q0 is smaller than α, and the difference between p1 and q0 and the difference between p0 and q1 is smaller than β. Mode 3 is a case where Bs is less than 4, and filters up to four pixels p1, p0, q0 and q1. Finally, mode 4 is a case where Bs is 4 and filters up to six pixels (p2, p1, p0, q0, q1, q2). α and β represent thresholds set according to quantization magnitudes as described in Table 1 below.

QPav 7 or less 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 α 0 4 4 5 6 7 9 10 12 14 17 20 24 28 33 39 β 0 3 3 3 4 4 4 6 6 7 7 8 8 9 9 10 QPav 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 α 46 55 65 76 90 106 126 148 175 207 245 255 255 255 255 255 255 β 10 11 11 12 12 13 13 14 14 15 15 16 16 17 17 18 18

Since mode 1 determines that the blocking phenomenon does not exist, the filtering process is not performed, and mode 2 processes only p0 and q0 values, and mode 3 is a default mode, and according to a specific condition, up to four pixels or It is supposed to process two pixels. On the other hand, mode 4 is determined to be a region where the blocking phenomenon considerably exists to process up to six pixels or at least two pixels according to a specific condition, and the filter coefficients are set differently from those in Table 1. Filtering is performed according to a formula determined for each mode by referring to the filter coefficients.

This process improves image quality, but has a problem in that it requires an excessive amount of computation. In other words, the deblocking filter that performs this filtering process in the decoding process has a large amount of calculation so that it takes 20 to 30% of the total processing time. In case of implementing such a deblock filter in hardware, since the filtering is performed in units of macroblocks, parallel processing cannot be performed. Meanwhile, the filtering process becomes complicated because different filtering is performed according to the condition determination. In addition, the ringing phenomenon (ringing artifact or flickering), which is one of compression loss phenomenon, appears in the process of eliminating temporal redundancy and has a problem that cannot be solved by the above filtering.

For the above reasons, there is a need for a fast filtering mechanism with a small amount of computation in video decoding.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-described necessity, and a technical object of the present invention is to provide a video decoding method having a fast filtering process and an apparatus therefor.

In addition, it is another technical problem to provide a method and an apparatus for eliminating the slump phenomenon.

In order to achieve the above object, the decoder for decoding an image encoded by the block-based video algorithm according to the present invention includes a filter unit for filtering the horizontal and vertical boundaries between blocks constituting the frame and temporally filtered the filtered frames. It is characterized by.

The filter unit is a spatial filter unit for correcting and filtering the values of the pixels by the block length around the boundary of the two blocks, and the temporal filter unit for performing the intermediate value filtering for a predetermined portion of the frames filtered by the spatial filter unit Include.

The spatial filter unit filters pixels corresponding to the block length when the difference between the block-to-block boundary pixel values is smaller than a predetermined value with respect to the boundary between blocks having a motion vector of a refiner unit of an intra-coded frame. The temporal filter unit filters the macroblocks of the preceding and following frames corresponding to the macroblocks with respect to macroblocks coded with a macroblock type, whose motion vector is smaller than a predetermined value, and whose quantization coefficient is smaller than a predetermined value. .

In order to achieve the above object, the H.264 decoder according to the present invention includes a loop filter unit and a post-processing filter unit and includes a filter selection unit to select any one of the loop filter unit and the post-processing filter unit. It is done.

The post-processing filter may be configured to filter the horizontal or vertical boundary between the sub-blocks constituting the frame by correcting four pixel values based on the horizontal or vertical boundary of the two sub-blocks. In this case, the post-processing filter unit may set the value of 4 when the difference between two pixel values abutting the boundary of the two subblocks with respect to the boundary between the subblocks having the motion vector of the refiner unit of the intra-coded frame is smaller than a predetermined value. It is preferable to correct the four pixel values according to the distributed form of the pixel values.

Meanwhile, the post-processing filter unit performs spatial filtering by correcting values of four pixels based on horizontal or vertical boundaries of the two subblocks in filtering horizontal or vertical boundaries between subblocks constituting the frame. Can be implemented to perform temporal filtering on these components. The temporal filtering performed by the post-processing filter unit corresponds to the macroblock for macroblocks coded in a macroblock type among the spatially filtered frames and whose motion vectors are smaller than a predetermined value and whose quantization coefficient is smaller than a predetermined value. Filter the macroblocks and the median of the before and after frames.

In order to achieve the above object, a video decoding method for decoding an image encoded by the block-based video algorithm according to the present invention includes filtering horizontal and vertical boundaries between blocks constituting a frame and temporally filtering the filtered frames. It is characterized by including.

The step may include: a spatial filtering step of correcting and filtering values of pixels by a block length around a boundary of two blocks; and a temporal filtering step of performing intermediate value filtering on a predetermined portion of frames filtered by the spatial filter unit. It includes.

The spatial filtering step filters the pixels of the block length when the difference in the inter-block boundary pixel value is smaller than a predetermined value with respect to the boundary between blocks having a motion vector of a refiner unit of an intra coded frame.

The temporal filtering step is to filter the macroblocks and the intermediate value of the preceding and following frames corresponding to the macroblocks for the macroblocks coded with the macroblock type, the motion vector is smaller than the predetermined value, and the quantization coefficient is smaller than the predetermined value. do.

In order to achieve the above object, the H.264 decoding method according to the present invention is characterized in that it can select any one of a loop filtering step of loop filtering the decoded image or a post-processing filtering step of post-processing the decoded image. .

The post-processing filtering step may be implemented to correct and filter values of four pixels around the horizontal or vertical boundary of the two sub-blocks in filtering the horizontal or vertical boundary between the sub-blocks constituting the frame. In this case, the post-processing filtering step may be performed when the difference between two pixel values that are in contact with the boundary of the two subblocks is smaller than a predetermined value with respect to the boundary between the subblocks having the motion vector of the refiner unit of the intra-coded frame. It is preferable to correct the four pixel values according to the distributed form of the pixel values.

In the post-processing filtering step, in filtering horizontal or vertical boundaries between sub-blocks constituting a frame, spatial filtering is performed by correcting values of four pixels around the horizontal or vertical boundaries of the two sub-blocks. It can be implemented to perform temporal filtering for. The temporal filtering is a macroblock of a front and back frame corresponding to the macroblock for macroblocks coded in a macroblock type among the spatially filtered frames and whose motion vectors are smaller than a predetermined value and whose quantization coefficient is smaller than a predetermined value. And median filtering.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Like reference numerals refer to like parts.

4 is a functional block diagram of an H.264 decoder according to an embodiment of the present invention.

The decoder of H.264 reconstructs the video image encoded by the H.264 encoder. First, the decoding process is performed by performing entropy decoding, and reconstructing the image through reordering, inverse quantization, and inverse transformation. The reconstructed image removes the unnaturalness due to blocking by the loop filter. Motion compensation is applied to a plurality of reference frames that are inter predicted using the motion vector, but does not undergo a motion compensation process for the intra predicted frame. H.264's entropy coding method uses CABAC (Context-based Adaptive Binary Arithmetic Coding) to remove probabilistic redundancy, and transform uses spatial transformation to remove spatial redundancy and motion compensation is 1/4. Perform up to pixels. The present invention is characterized in that the filtering process is improved in video decoding through a filtering process like the H.264 loop filter.

Meanwhile, in the detailed description of the present invention, an embodiment is described based on H.264, but the technical spirit of the present invention is not limited thereto. The part improved by the present invention in FIG. 4 is the filter part 100, and the detailed structure of the filter part will be described with reference to FIGS. 5A, 5B and 5C.

5A, 5B, and 5C are functional block diagrams showing the filter part of FIG. 4 in more detail, respectively.

The filter unit 100 may be implemented to be compatible with an existing decoding process.

The filter unit 100 of FIG. 5A allows the filter selector 110 to select a filter existing in an existing decoder and a filter newly provided by the present invention. That is, the filter selector 110 may select one of the space and the temporal filter 120 and the H.264 filter 130 newly provided by the present invention by the user's setting or the like, and FIG. 5A shows the space. And the temporal filter unit 120 is selected. The image input to the filter unit 100 is filtered and output through the spatial and temporal filter unit 120.

The filter unit 100 of FIG. 5B allows the existing H.264 filter unit 130 to be used as in FIG. 5A. In the case of the H.264 filter unit 130, only the unnaturalness due to the block is removed and the filtering cannot be performed on temporal fluctuations. Thus, the embodiment illustrated in FIG. 5B passes through the H.264 filter unit 130. A temporal filter unit 150 is provided to reduce temporal blurring of an image. Meanwhile, when the spatial filter unit 140 is selected as shown by the filter selecting unit 110, the spatial filter unit 140 reduces the unnaturalness due to the blocking of the image and temporally through the temporal filter unit 150. You can get rid of slack.

The filter unit 100 of FIG. 5C uses the existing H.264 filter unit 130 to reduce the unnaturalness due to the block, and reduces the temporal rung of the image filtered through the H.264 filter unit 130. The unit 160 outputs an image from which temporal rung is removed.

5A to 5C are all implemented to satisfy the existing decoder standard, and may perform temporal filtering that the conventional decoder could not perform. In the case of FIG. 5A or 5B, the H.264 filter unit 130 is used when the loop filter is required in the encoding process and the loop filter is required in the decoding process. Spatial filtering can be performed to significantly reduce the unnaturalness caused by blocking. On the other hand, Figures 5a to 5c can all eliminate the temporal haze. A preferred embodiment of the process of reducing unnaturalness due to blocking will be described with reference to FIGS. 6A and 6B, and a process of reducing temporal rung will be described with reference to FIGS. 7 and 8.

6A shows samples between adjacent 4 * 4 subblocks subject to spatial filtering, and FIG. 6B shows a case of spatial filtering separated according to a preferred embodiment.

FIG. 6A shows two 4 * 4 subblocks adjacent in the horizontal direction.

The filtering is performed based on a boundary of two subblocks based on a maximum of four pixels to prevent overlapping filtering between subblocks so that the filtering process can be parallelized. On the other hand, if the image is filtered by an algorithm that compresses a video by obtaining a motion vector based on 8 * 8 blocks, filtering between blocks does not overlap each other when filtering based on a maximum of 8 pixels, and thus the filtering process is parallel. Can be processed as That is, when filtering on the boundary of two blocks, it is preferable to filter based on the pixels of the block length. In a preferred embodiment of the present invention, it is desirable to reduce unnaturalness due to blocking only in intra-coded frames in order to reduce the amount of computation. In the case of an inter frame, prediction is obtained from an image of a previous frame, because interpolation is performed at this time. Since H.264's motion compensation is performed up to 1/4 pixel, it also filters 1/2 and 1/4 pixels. For a half-pixel, use a six-tap filter of (1, -5, 20, 20, -5, 1), and for a quarter-pixel, use the average of half-pixels. Therefore, since the reference image made for motion compensation has already performed the filtering process, the deblocking filtering process that reduces the unnaturalness of the blocking is not necessary unless the motion vector of the refinery unit is used. Hereinafter, spatial filtering (deblocking filtering) will be described. The filtering is performed in the horizontal direction and the vertical direction. Since the filtering in the horizontal direction and the vertical direction is performed almost similarly, only the horizontal direction will be described in the embodiment. The filtering may be divided into a step of determining whether to perform filtering first and a step of performing actual filtering. Referring to FIG. 6A, whether to perform filtering is determined based on a difference between B and C pixel values and a quantization value. That is, if the difference between B and C is smaller than a certain threshold, filtering is performed. Otherwise, filtering is not performed. See Table 2 for the criteria for determining whether to perform filtering.

QP (quantization value) 1-19 20-25 26-31 32-36 37-44 45-50 Reference value Do not perform filtering 5 10 20 40 50

When the difference between two pixels B and C is smaller than the reference value, the reason for performing filtering is that when the difference between two pixels is large, it can be judged as the difference in the original image rather than the difference due to blocking, and in this case, the filtering is performed. It is proper not to do it. On the other hand, since deblocking filtering is a process of reducing unnaturalness due to blocking, it performs a function as a kind of lowpass filter, so that the characteristics of the filter can be modeled by Equation 1.

However, rather than calculating deblocking filtering using Equation 1, Equation 1 is approximated to reduce the amount of computation. 6B shows four cases of deblocking filtering. The four cases show the case where the value of pixel C is larger than the value of pixel B, and the same can be considered when the value of pixel C is smaller than the value of pixel B.

First, cases 1 to 4 may be approximated by Equations 2 to 5, respectively.

A '= A + (D-A) / 16, B' = B + (D-A) / 4, C '= C- (D-A) / 4, D' = D- (D-A) / 16

A '= A, B' = A + (D-A) / 4, C '= D- (D-A) / 4, D' = D

A '= A + (D-A) / 16, B' = B + (D-A) / 4, C '= D- (D-A) / 4, D' = D

A '= A, B' = A + (D-A) / 4, C '= C- (D-A) / 4, D' = D- (D-A) / 16

Here, A ', B', C ', and D' denote values of respective pixels after filtering, and A, B, C, and D denote values of respective pixels before filtering.

Referring to Case 1 with reference to Equation 2, the values are increasing in the order of A, B, C, and D. As the gap between B and C is reduced, blocking is reduced, which causes A and D to be wider with B and C, respectively. To alleviate this, A is smaller than B but increases slightly, while D is smaller than C but slightly lower. This ensures a smooth overall connection. In Equation 2, the correction degree of B and C at the boundary is four times the correction degree of A and D.

Referring to Case 2 with reference to Equation 3, A and D are not significantly separated from the center, so in this case, the values are left as they are and B and C are corrected. At this time, B and C are corrected by using A and D close to their centers as reference values, respectively.

Referring to Case 3 with reference to Equation 4, since D has a value close to the center, it is left as it is and the remaining A, B, and C values are corrected. A and B values are corrected as in case 1, and in C, as in case 2.

Referring to Case 4 with reference to Equation 5, A and B are treated like Case 2 and C and D are processed as Case 1.

If the value of B is larger than the value of C, deblocking filtering may be performed in the same manner as in Cases 1 to 4. After deblocking filtering, temporal filtering is performed to reduce temporal flickering. Temporal wobble is noticeable when appearing in the background area, which results in a deterioration of the image quality. Therefore, it is necessary to determine whether the image is a background area or a motion area. Typical motion detection is based on the difference between two consecutive frames. Differences between successive frames appear when there is motion or noise. To detect motion, Philips's method finds the difference between two consecutive frames and passes the lowpass filter twice. Since such a motion detection method requires a large amount of computation, a process of detecting using a motion vector and a block type and performing temporal filtering through motion detection will be described with reference to FIGS. 7 and 8.

7 is a diagram illustrating an example of a differential image that undergoes motion compensation.

The image of FIG. 7 shows a difference image obtained by obtaining a difference between images after motion compensation. As shown, it can be seen that the background image is made of 16 * 16 macroblocks and the size of the difference image is very small. That is, it can be seen that the background part has a large block size that is a unit of motion, and the difference image has a very small value. Therefore, if the largest block type and the motion vector value is smaller than a predetermined reference value, the temporal filtering is performed. In the case of an H.264 image, it is preferable to limit the quantization coefficient to a 16 * 16 macroblock type with a motion vector less than 3 and a quantization coefficient of 10 or less. If this condition is satisfied, temporal filtering is performed. The median filter is used here.

8 is a diagram illustrating adjacent frames that are subjected to temporal filtering according to an embodiment of the present invention. As shown in FIG. 8, temporal filtering for a macroblock satisfying the temporal filtering condition is determined as an intermediate value between three frames.

Those skilled in the art will appreciate that the present invention can be embodied in other specific forms without changing the technical spirit or essential features of the present invention. For example, although an embodiment has been described based on H.264, the present invention can also be applied to decoding of an image using another video compression algorithm. Therefore, it should be understood that the embodiments described above are exemplary in all respects and not restrictive. The scope of the present invention is indicated by the scope of the following claims rather than the detailed description, and all changes or modifications derived from the meaning and scope of the claims and the equivalent concept are included in the scope of the present invention. Should be interpreted.

The present invention provides a filter capable of removing unnaturalness due to blocking with a small amount of calculation. The filter has a feature that can be used with a conventional decoding method. Meanwhile, according to the present invention, since the temporal filter for temporal filtering may be used together with the deblocking filter for performing spatial filtering, the present invention has an advantage of reducing temporal flickering of the video.

1 shows a boundary of a macroblock to be filtered.

2 is a diagram illustrating a process of determining a block size according to a conventional technique.

3 shows samples in adjacent 4 * 4 subblocks in a horizontal or vertical direction.

4 is a functional block diagram of an H.264 decoder according to an embodiment of the present invention.

5A, 5B, and 5C are functional block diagrams showing the filter part of FIG. 4 in more detail, respectively.

FIG. 6A illustrates samples between adjacent 4 * 4 subblocks subject to spatial filtering according to an embodiment of the present invention.

6B illustrates a case of spatial filtering according to an embodiment of the present invention.

7 is a diagram illustrating an example of a differential image that undergoes motion compensation.

8 is a diagram illustrating adjacent frames that are subjected to temporal filtering according to an embodiment of the present invention.

Claims (18)

A decoder for decoding an image encoded with a block-based video algorithm, A video decoder comprising a filter unit for filtering horizontal and vertical boundaries between blocks constituting a frame and temporally filtering the filtered frames. The method of claim 1, wherein the filter unit A spatial filter unit for correcting and filtering values of pixels corresponding to block lengths based on a boundary between two blocks; And A temporal filter unit performing intermediate value filtering on a predetermined portion of the frames filtered by the spatial filter unit; Video decoder comprising a 3. The pixel filter of claim 2, wherein the spatial filter filters the pixels of the block length when the difference between the block-to-block boundary pixel values is smaller than a predetermined value with respect to the boundary between blocks having a motion vector of a refiner unit of an intra-coded frame. Video decoder, characterized in that The macroblock of claim 2, wherein the temporal filter unit is coded in a macroblock type, the motion vector is smaller than a predetermined value, and the quantization coefficient is smaller than the predetermined value. Video decoder characterized in that the median filtering with A H.264 decoder for decoding an encoded image, An H.264 decoder comprising a loop filter unit and a post-processing filter unit for filtering the decoded image, and a filter selection unit for selecting one of the loop filter unit and the post-processing filter unit. The method of claim 5, wherein the post-processing filter unit In filtering horizontal or vertical boundaries between sub-blocks constituting a frame, the H.264 decoder is characterized by correcting and filtering four pixel values based on the horizontal or vertical boundaries of the two sub-blocks. 7. The method of claim 6, wherein the post-processing filter unit is less than a predetermined value when a difference between two pixel values in contact with the boundary of the two subblocks with respect to the boundary between the subblocks having the motion vector of the refiner unit of the intra-coded frame is less than a predetermined value. H.264 decoder, characterized in that for correcting the four pixel values in accordance with the distributed form of the four pixel values in 6. The method of claim 5, wherein the post-processing filter unit filters the horizontal or vertical boundaries between the sub-blocks constituting the frame, and spatially filters the four pixels by correcting values of the four pixels around the horizontal or vertical boundaries of the two sub-blocks. H.264 decoder characterized by performing temporal filtering on spatially filtered frames The method of claim 8, wherein the temporal filtering performed by the post-processing filter is performed on macroblocks of the spatially filtered frame, coded in a macroblock type, having a motion vector smaller than a predetermined value, and having a quantization coefficient smaller than a predetermined value. H.264 decoder characterized in that the intermediate value filtering with the macroblocks of the front and back frames corresponding to the macroblocks. In the video decoding method for decoding an image encoded with a block-based video algorithm, Filtering horizontal and vertical boundaries between blocks constituting the frame, and temporally filtering the filtered frames. 11. The method of claim 10, wherein the step comprises: a spatial filtering step of correcting and filtering values of pixels by a block length around a boundary of two blocks; And A temporal filtering step of performing intermediate value filtering on a predetermined portion of the frames filtered by the spatial filter unit; Video decoding method comprising a 12. The method of claim 11, wherein the spatial filtering step selects pixels equal to the block length when the difference in the inter-block boundary pixel value is smaller than a predetermined value with respect to the inter-block boundary having a motion vector of a refiner unit of an intra coded frame. Video decoding method characterized in that the filtering 12. The method of claim 11, wherein the temporal filtering step includes macros of preceding and following frames corresponding to the macroblocks for macroblocks coded with a macroblock type, motion vectors smaller than a predetermined value, and quantization coefficients smaller than a predetermined value. Video decoder characterized in that the median filtering with blocks In the H.264 decoding method for decoding an encoded image, H.264 decoding method comprising the step of selecting a loop filtering step of loop filtering the decoded image or a post-processing filtering step of post-processing the decoded image. 15. The method of claim 14, wherein the post-processing filtering step comprises filtering four pixels by correcting values of horizontal or vertical boundaries between the two subblocks in filtering horizontal or vertical boundaries between subblocks constituting the frame. H.264 decoding method 16. The method of claim 15, wherein the post-processing filtering step includes a difference between two pixel values abutting a boundary between the two subblocks with respect to a boundary between subblocks having a motion vector of a refiner unit of an intra coded frame is smaller than a predetermined value. H.264 decoding method, characterized in that for correcting the four pixel values in accordance with the distribution form of the four pixel values 15. The method of claim 14, wherein the post-processing filtering step performs spatial filtering by correcting values of four pixels around the horizontal or vertical boundary of the two subblocks in filtering the horizontal or vertical boundary between the subblocks constituting the frame. H.264 decoding method for performing temporal filtering on the spatially filtered frames 18. The method of claim 17, wherein the temporal filtering corresponds to the macroblock for macroblocks of the spatially filtered frame that are coded in a macroblock type and whose motion vector is less than a predetermined value and whose quantization coefficient is less than a predetermined value. H.264 decoding method characterized in that the median filtering with macroblocks of the frame before and after