USRE43628E1 - Method and apparatus for controlling loop filtering or post filtering in block based motion compensationed video coding - Google Patents

Abstract

Adjacent blocks are identified in an image. Coding parameters for the adjacent blocks are identified. Deblock filtering between the identified adjacent blocks is skipped if the coding parameters for the identified adjacent blocks are similar and not skipped if the coding parameters for the identified adjacent blocks are substantially different.

Description

This application is a divisional based on prior U.S. Ser. No. 09/817,701, filed Mar. 26, 2001 now U.S. Pat. No. 6,931,063.

BACKGROUND

Block based motion compensated video coding is used in many video compression standards such as H.261, H.263, H263+, MPEG-1, MPEG-2, and H26L. The lossy compression process can create visual artifacts in the decoded images, referred to as image artifacts. Blocking artifacts occur along the block boundaries in an image and are caused by the coarse quantization of transform coefficients.

Image filtering techniques can be used to reduce artifacts in reconstructed images. Reconstructed images are the images produced after being inverse transformed and decoded. The rule of thumb in these techniques is that image edges should be preserved while the rest of the image is smoothed. Low pass filters are carefully chosen based on the characteristic of a particular pixel or set of pixels surrounding the image edges.

Non-correlated image pixels that extend across image block boundaries are specifically filtered to reduce blocking artifacts. However, this filtering can introduce blurring artifacts into the image. If there are little or no blocking artifacts between adjacent blocks, then low pass filtering needlessly incorporates blurring into the image while at the same time wasting processing resources.

The present invention addresses this and other problems associated with the prior art.

SUMMARY OF THE INVENTION

Adjacent blocks are identified in an image. Coding parameters for the adjacent blocks are identified. Deblock filtering between the identified adjacent blocks is skipped if the coding parameters for the identified adjacent blocks are similar and not skipped if the coding parameters for the identified adjacent blocks are substantially different.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a diagram showing how deblock filtering is selectively skipped according to similarities between adjacent image blocks.

FIG. 2 is a diagram showing two adjacent image blocks having similar motion vectors.

FIG. 3 is a diagram showing how transform coefficients are identified for one of the image blocks.

FIG. 4 is a diagram showing how residual transform coefficients are compared between two adjacent image blocks.

FIG. 5 is a block diagram showing how the video image is encoded and decoded.

FIG. 6 is a block diagram showing how deblock filtering is selectively skipped in a codec.

FIG. 7 shows a table containing the results from selective deblock filter skipping.

DETAILED DESCRIPTION

In conventional filtering methods, filter processing only considers a single reconstructed image frame at a time. The motion-vector information available at both the encoder and decoder is not used. If two adjacent blocks share the same motion vector with respect to the same reference image frame, (for a multiple reference frames system) there may be no significant difference between the image residuals of each block. The block boundary of these two adjacent blocks may have been filtered in the reference frame and should therefore not be filtered again for the current frame. If a deblock filter is used without considering this motion-vector information, the conventional filtering process might filter the same boundary again and again from frame to frame. This unnecessary filtering not only causes unnecessary blurring but also results in extra filter computations.

FIG. 1 shows animage12 that selectively filters block artifacts according to similarities between image blocks. Theimage12 includesmultiple image blocks14 that are each individually encoded before being stored, transmitted, etc. The borders between some of theblocks14 includeblocking artifacts18. Blocking artifacts are any image discontinuities betweenblocks14 that may be created by the encoding process. A low pass filter is used to reduce the blocking artifacts that exist at the borders of adjacent image blocks.

For example, blockingartifacts24 exist betweenblocks20 and22. A low pass filter is used at theborder26 betweenblocks20 and22 to remove or reduce theblocking artifacts24. The low pass filter in one example selects a group ofpixels28 from both sides of theborder26. An average pixel value is derived from the group ofpixels28. Then each individual pixel is compared to the average pixel value. Any pixels ingroup28 outside of a predetermined range of the average pixel value is then replaced with the average pixel value.

As described above, if there are little or noblocking artifacts24 between the adjacent pixels, then the group ofpixels28 may be needlessly filtered causing blurring in the image and wasting processing resources. A skip mode filtering scheme uses the motion estimation and compensation information for adjacent image blocks. If the motion estimation and compensation information is similar, deblock filtering is skipped. This not only avoids unnecessary image blurring but also significantly reduces the required number of filtering operations.

For example, it is determined during the encoding process that adjacent image blocks30 and32 have similar coding parameters. Accordingly, deblock filtering is skipped for the groups ofpixels34 that extend across theborder31 betweenadjacent blocks30 and32. Skip mode filtering can be used for any horizontal or vertical boundary between any adjacent blocks inimage12.

FIG. 2 showsreference frames42 and48 and acurrent frame40 that is currently being encoded or decoded. Coding parameters forblocks44 and46 are compared to determine whether deblock filtering should be skipped between the twoadjacent blocks44 and46. One encoding parameter that is compared is the Motion Vectors (MV) for theblocks44 and46.

The motion vector MV1 points fromblock44 incurrent image frame40 to an associatedblock44′ in thereference image42. The motion vector MV2 points fromblock46 incurrent image frame40 to an associatedblock46′ inreference frame42. Skip mode filtering checks to see if the motion vectors MV1 and MV2 point to adjacent blocks in thesame reference frame42. If the motion vectors point to adjacent blocks in the reference frame (MV1=MV2), then deblock filtering is skipped. This motion vector information may be used along with other coding information to decide whether to skip deblock filtering between the twoimage blocks44 and46.

More than one reference frame may be used during the encoding and decoding process. For example, there may be anotherreference frame48. Theadjacent blocks44 and46 may have motion vectors pointing to different reference frames. In one embodiment, the decision to skip deblock filtering depends on whether the motion vectors for the two adjacent blocks point to the same reference frame. For example,image block44 may have amotion vector49 pointing toreference frame48 andimage block46 may have the motion vector MV2 pointing toreference frame42. Deblock filtering is not skipped in this example because themotion vectors49 and MV2 point to different reference frames.

FIG. 3 shows another coding parameter that may be used to decide whether or not to skip deblock filtering. Theimage block44 fromimage frame40 is compared withreference block44′ from thereference frame42 pointed to by the motion vector MV1 as previously shown inFIG. 2. Aresidual block44″ is output from the comparison betweenimage block44 andreference block44′. Atransform50 is performed on theresidual block44″ creating a transformedblock44′″ of transform coefficients. In one example, thetransform50 is a Discrete Cosine Transform. The transformedblock44′″ includes aD.C. component52 andA.C. components53.

TheD.C. component52 refers to a lowest frequency transform coefficient inimage block44. For example, the coefficient that represents the average energy in theimage block44. TheA.C. components53 refer to the transform coefficients that represent the higher frequency components in theimage block44. For example, the transform coefficients that represent the large energy differences between pixels in theimage block44.

FIG. 4 shows the transformedresidual blocks44′″ and46′″. TheD.C. components52 from the two transformedblocks44′″ and46′″ are compared inprocessor54. If the D.C. components are the same or within some range of each other, theprocessor54 notifies adeblock filter operation56 to skip deblock filtering between the border of the twoadjacent blocks44 and46. If theD.C. components52 are not similar, then no skip notification is initiated and the border betweenblocks44 and46 is deblock filtered.

In one embodiment, skip mode filtering is incorporated into the Telecommunications Sector of the International Telecommunication Union (ITU-T) proposed H.26L encoding scheme. The H.26L scheme only uses 4×4 integer Discrete Cosine Transform (DCT) blocks. Here, only the D.C. component of the two adjacent blocks may be checked. However some limited low frequency A.C. coefficients could also be checked when the image blocks are bigger sizes, such as 8×8 or 16×16 blocks. For example, theupper D.C. component52 and the three lower frequency A.C. transformcoefficients53 forblock44′″ may be compared with theupper D.C. component52 and three lower frequency A.C. transformcoefficients53 forblock46′″. Different combinations of D.C. and/or low frequency A.C. transform coefficients can be used to identify the relative similarity between the twoadjacent blocks44 and46.

Theprocessor54 can also receiveother coding parameters55 that are generated during the coding process. These coding parameters include the motion vectors and reference frame information for theadjacent blocks44 and46 as described above. Theprocessor54 uses all of these coding parameters to determine whether or not to skip deblock filtering between adjacent image blocks44 and46. Other encoding and transform functions performed on the image may be carried out in thesame processor54 or in a different processing circuit. In the case where all or most of the coding is done in the same processor, the skip mode is simply enabled by setting a skip parameter in the filtering routine.

FIG. 5 shows how skip mode filtering is used in a block-based motion-compensated Coder-Decoder (Codec)60. Thecodec60 is used for inter-frame coding. An input video block from the current frame is fed frombox62 into acomparator64. The output of aframe buffering box80 generates areference block81 according to the estimated motion vector (and possible reference frame number). The difference between the input video block and thereference block81 is transformed inbox66 and then quantized inbox68. The quantized transform block is encoded by a Variable Length Coder (VLC) inbox70 and then transmitted, stored, etc.

The encoding section of thecodec60 reconstructs the transformed and quantized image by first Inverse Quantizing (IQ) the transformed image inbox72. The inverse quantized image is then inverse transformed inbox74 to generate a reconstructed residual image. This reconstructed residual block is then added inbox76 to thereference block81 to generate a reconstructed image block. Generally the reconstructed image is loop filtered inbox78 to reduce blocking artifacts caused by the quantization and transform process. The filtered image is then buffered inbox80 to form reference frames. The frame buffering inbox80 uses the reconstructed reference frames for motion estimation and compensation. Thereference block81 is compared to the input video block incomparator64. An encoded image is output atnode71 from the encoding section and is then either stored or transmitted.

In a decoder portion of thecodec60, a variable length decoder (VLD) decodes the encoded image inbox82. The decoded image is inverse quantized inbox84 and inverse transformed inbox86. The reconstructed residual image frombox86 is added in the summingbox88 to thereference block91 before being loop filtered inbox90 to reduce blocking artifacts and buffered inbox92 as reference frames. Thereference block91 is generated frombox92 according to the received motion vector information. The loop filtered output frombox90 can optionally be post filtered inbox94 to further reduce image artifacts before being displayed as a video image inbox96. The skip mode filtering scheme can be performed in any combination of the filtering functions inboxes78,90 and94.

The motion estimation and compensation information available during video coding are used to determine when to skip deblock filtering inboxes78,90 and/or94. Since these coding parameters are already generated during the encoding and decoding process, there are no additional coding parameters that have to be generated or transmitted specially for skip mode filtering.

FIG. 6 shows is further detail how skip mode filtering is used in thefilters78,90, and/or94 in the encoder and decoder inFIG. 5. The interblock boundary between any two adjacent blocks “j” and “k” is first identified inbox100. The two blocks may be horizontally or vertically adjacent in the image frame.Decision box102 compares the motion vector mv(j) for block j with the motion vector mv(k) for block k. It is first determined whether the two adjacent blocks j and k have the same motion vector pointing to the same reference frame. In other words, the motion vectors for the adjacent blocks point to adjacent blocks (mv(j)=mv(k)) in the same reference frame (ref(j)=ref(k)).

It is then determined whether the residual coefficients for the two adjacent blocks are similar. If there is no significant difference between the image residuals of the adjacent blocks, for example, the two blocks j and k have the same of similar D.C. component (dc(j)=dc(k)), then the deblock filtering process inbox104 is skipped. Skip mode filtering then moves to the next interblock boundary inbox106 and conducts the next comparison indecision box102. Skip mode filtering can be performed for both horizontally adjacent blocks and vertically adjacent blocks.

In one embodiment, only the reference frame and motion vector information for the adjacent image blocks are used to determine block skipping. In another embodiment, only the D.C. and/or A.C. residual coefficients are used to determine block skipping. In another embodiment, the motion vector, reference frame and residual coefficients are all used to determine block skipping.

The skip mode filtering scheme can be applied to spatially sub-sampled chrominance channels For example in a case with 4:2:0 color format sequences, skip mode filtering for block boundaries may only rely on the equality of motion vectors and D.C. components for the luminance component of the image. If the motion vectors and the D.C. components are the same, deblock filtering is skipped for both the luminance and chrominance components of the adjacent image blocks. In another embodiment, the motion vectors and the D.C. components are considered separately for each luminance and chrominance component of the adjacent blocks. In this case, a luminance or chrominance component for adjacent blocks may be deblock filtered while the other luminance or chrominance components for the same adjacent blocks are not deblock filtered.

FIG. 7 includes a table110 showing the results of skip mode filtering using a ITU-TH.26L Testing Model-Long TML5.0. Table110 compares the results of the TML filtering standard with skip mode filtering as described above. Encoding results using skip mode filtering are shown in table110 under the heading SLA.

There were four images that were tested, Akiyo_cif for 300 frames at 30 Frames Per Second (fps), Foreman_cif for 300 frames at 30 fps, Foreman_qcif for 100 frames at 10 fps, and Tempete_cif for 260 frames at 30 fps. Quantization Parameters (QP) of 25 and 30 were used. The results show no significant visual quality degradation with the skip mode filtering. The Picture Signal to Noise Ratio (PSNR) for the images stays approximately the same for the luminance Y and chrominance U and V channels. However, skip mode filtering provides time savings of 40-70 percent.

Skip mode filtering can be used with any system that encodes or decodes multiple image frames. For example, DVD players, video recorders, or any system that transmits image data over a communications channel, such as over television channels or over the Internet.

The skip mode filtering described above can be implemented with dedicated processor systems, micro controllers, programmable logic devices, or microprocessors that perform some or all of the operations. Some of the operations described above may be implemented in software and other operations may be implemented in hardware.

For the sake of convenience, the operations are described as various interconnected functional blocks or distinct software modules. This is not necessary, however, and there may be cases where these functional blocks or modules are equivalently aggregated into a single logic device, program or operation with unclear boundaries. In any event, the functional blocks and software modules or described features can be implemented by themselves, or in combination with other operations in either hardware or software.

Having described and illustrated the principles of the invention in a preferred embodiment thereof, it should be apparent that the invention may be modified in arrangement and detail without departing from such principles. Claim is made to all modifications and variation coming within the spirit and scope of the following claims.

Claims (17)

1. A method for encoding an image, comprising:

identifying spatially adjacent blocks in a same frame of the image;

identifying transform coefficients motion vectors used for motion compensation for the spatially adjacent blocks in the same frame of the image and associated reference frames for the motion vectors of the spatially adjacent blocks;

comparing the transform coefficients betweenjudging whether to perform deblock filtering for removing image residuals between the identified spatially adjacent block caused by encoding the image, by: determining whether the motion vectors for the spatially adjacent blocks in the same frame of the image are the same or similar to each other and determining whether the reference frames for motion vectors of the spatially adjacent blocks are identical; and

skipping deblock filtering for removing image residuals caused by encoding the image between the identified spatially adjacent blocks when the comparison indicates that the motion vectors for the spatially adjacent blocks in the same frame of the image have are determined to be the same or similar transform coefficients; and

deblock filtering to remove image residuals between the identified spatially adjacent blocks when the comparison between the transform coefficients for the identified spatially adjacent blocks in the same frame of the image indicate that the spatially adjacent blocks do not have the same or similar transform coefficients and the reference frames for the motion vectors of the spatially adjacent blocks are determined to be identical.

2. A method according toclaim 1 including:

identifying D.C. components in the transform coefficients for the spatially adjacent blocks;

comparing the D.C. components for the spatially adjacent blocks with each other; and

skipping deblock filtering between the spatially adjacent blocks when the D.C. components for the compared spatially adjacent blocks have same or similar values relative to each other.

3. A method according toclaim 1 including:

identifying both D.C. and A.C. components in the transform coefficients for the spatially adjacent blocks;

comparing the D.C. components for adjacent blocks and A.C. components for adjacent blocks; and

skipping deblock filtering between the adjacent blocks when both the D.C. components are the same or similar and the A.C. components are the same or similar.

4. A method according toclaim 1 including controlling deblock filtering for a loop filter in an image coder.

5. A method according toclaim 1 including controlling deblock filtering in one or both of a loop filter and a post filter in an image decoder.

6. A method according toclaim 1 including:

identifying similarities between coding parameters in a luminance channel of the adjacent blocks; and

controlling deblock filtering for both the luminance channel and a chrominance channel in the image according to identified similarities in the luminance channel.

7. A method according toclaim 1 including selectively skipping deblock filtering in any one of a H.261, H.263, H263+, MPEG-1, MPEG-2, or H26L encoding standard according to encoding parameter similarities between adjacent image blocks.

8. An encoder for encoding an image, comprising:

a processor adapted to:

compare blocks in a same image frame with reference frames; transform a result of the comparison between the reference frames and the blocks in the image frame into transformed blocks having transform coefficients;identify motion vectors and associated reference frames for the motion vectors of spatially adjacent blocks in a same frame of the image;

compare are the similarities between the transform coefficientsdetermine whether motion vectors for spatially adjacent transformed blocks within the same image frame are the same or similar to each other and determine whether reference frames for the motion vectors of the spatially adjacent blocks are identical; and

skippingskip deblock filtering between spatially adjacent transformed blocks in the image that have the same or similar transform coefficients motion vectors and that have identical reference frames for the motion vectors.

9. An encoder according toclaim 8 wherein the processor is adapted to

identify motion vectors and associated reference frames for the adjacent blocks and skip deblock filtering between the spatially adjacent blocks according to the identified motion vectors and reference frames.

10. An encoder according toclaim 8 wherein the processor is adapted to identify residual coefficients for the spatially adjacent blocks and skip deblock filtering according to the identified residual coefficients.

11. An encoder according toclaim 8 wherein the processor is adapted to skip deblock filtering according to both D.C. and A.C. components in the residual coefficients.

12. A decoder for decoding an encoded image, comprising:

a processor adapted to identify spatially adjacent blocks in the encoded image;

identify coding parameters including D.C. coefficients motion vectors and associated reference frames for the motion vectors of the spatially adjacent blocks within a same encoded image frame;

compare the motion vectors and associated references frames for the motion vectors of the identified D.C. coefficients between coding parameters of the identified spatially adjacent blocks; and

enable or disable filtering of blocking artifacts between the spatially adjacent blocks according to the comparison of the D.C. coefficients coding parameters between the spatially adjacent blocks within the same encoded image frame, wherein

the processor skips deblock filtering when the motion vectors for the spatially adjacent blocks are the same or similar and the reference frames for motion vectors of the spatially adjacent blocks are identical.

13. A decoder according toclaim 12 wherein the processor is adapted to

identify both D.C. coefficients and A.C. coefficients for the spatially adjacent blocks within a same encoded image frame;

compare the identified D.C. coefficients between the identified spatially adjacent blocks and corn are compare the identified A.C. coefficients between the identified spatially adjacent blocks; and

enable or disable filtering of blocking artifacts between the spatially adjacent blocks according to the comparison of the D.C. coefficients between the adjacent blocks within the same encoded image frame and according to the comparison of the A.C. coefficients between the adjacent blocks within the same encoded image frame.

14. A decoder according toclaim 12 wherein the processor is adapted to identify residual coefficients for the spatially adjacent blocks and skip deblock filtering between the spatially adjacent blocks according to the identified residual coefficients.

15. A decoder according toclaim 12 wherein the processor is adapted to

inverse transform the encoded image,

compare blocks in the inverse transformed encoded image with reference frames,

generate a reconstructed image from the comparison between the inverse transformed encoded image and the reference frame, and

skip deblock filtering between spatially adjacent blocks in the reconstructed image according to the coding parameters for the spatially adjacent blocks.

16. A decoder according toclaim 15 wherein the coding parameters include transform coefficients, motion vectors, and reference frame information and the processor skips deblock filtering when the transform coefficients, motion vectors, and reference frame information for the spatially adjacent blocks are the same or similar.

17. A decoder according toclaim 12 wherein the processor is adapted to skip deblock filtering in one or both of a loop filter and a post filter.

Images