Joint Video Experts Team (jvet) of itu-t sg 6 wp and iso/iec jtc 1/sc 29/wg 11

7.4Loop filters (4)

Contributions in this category were discussed Saturday 14 April 1730–XXXX (chaired by JRO and GJS).

JVET-J0038 Signal Adaptive Diffusion Filters for Video Coding [J. Pfaff, J. Rasch, M. Schäfer, H. Schwarz, M. Winken, A. Henkel, M. Siekmann, D. Marpe, T. Wiegand (HHI)]

In this documentcontribution, diffusion filters are introduced proposed that may be applied to the prediction signal generated by a hybrid video codec. Two types of diffusion filters are proposed: Linear and nonlinear diffusion filters. The linear diffusion filters correlate the extended prediction signal a number ofn times using a symmetric filter mask. The nonlinear diffusion filters use the input prediction signal to identify structures of the underlying signal and diffuse along edges rather than perpendicular to them. It wais reported that the proposed diffusion filters lead to a codinghas a bit-rate delta of up to −1.45% in the All IntraAI configuration and −2.00% in Random Accessthe RA configuration.

Average gains weare reported for HD and UHD sequences (mixture of CTC class A/B and CfP sequences). In RA, the reported bit rate reduction is 1.09% for high rates (QP values of 22..37), and 0.65% for lower bit rates. The gGain seems to be higher for higher resolutions.

NThis is not a loop filter that operates at the end of the decoded pictures, but rather is applied in the prediction signal generation process.

Up to five different configurations of the filter are applied, depending on the block size and temporal layer.

The mMaximum number of iterations is 35 in the linear case and, 20 in the nonlinear case (where each of the iterations is more complex).

This dDoes not have an impact on additional memory access, but the number of operations is clearly higher than for interpolation filtering.

For further studyFurther study of this was requested.

JVET-J0056 Multi-Dimensional Filter Selection for Deblocking [J. Dong, Y.-H. Chao, W.-J. Chien, L. Zhang, M. Karczewicz (Qualcomm)]

This contribution presents a multi-dimensional filter selection scheme for deblocking. The filter selection for a sample is four-dimensional, i.e., determined by four factsaspects: the average local activities of two bBlocks P and Q, the difference of local activities of bBlocks P and Q, the list type of the belonged block (tType 0 or tType 1), and the distance from the segment. Given a combination of the four factselements, the filter index is not fixed, but is adaptively determined by the encoder and is signalled in the bitstream. This contribution leverages uses a bit more computation resource for significant coding efficiency improvement, while still reportedly being easy for to implement using parallel processing.

The reported bBit rate reduction is 1.25% for AI and 1.41% for RA in the CTC. This is however with all tools except QTBT disabled.

It was asked how the performance would be with ALF enabled. It was nNot known how it would perform with other tools on.

The eEncoder runtime is was not changed, but the decoder runtime increases by 18% in AI, 10% in RA.

The aActivity criterion is based on a seco2nd derivative.

15 different filter types (predefined) filter types are used.

A lLookup table that determines which filter to select through based the 4 criteria is determined at the encoder side and transmitted. The encoder designs the lookup table after encoding and /decoding the frame, and determines which of the filters optimize the reconstruction locally, and designs the LUT based on that.

More information was needed how it interacts with other tools.

JVET-J0071 Non-local Structure-based Filter with integer operation [X. Meng, C. Jia, Z. Wang, S. S. Wang, S. Ma (Peking University), X. Zheng (DJI)]

This contribution is a continuation ofbased on the NLSF (nNonlocal sStructure in-loop fFilter) technique that was proposed in JVET-J0011. It proposes a solutionan approach for an integer NLSF algorithm. The NLSF design in JVET-J0011 contains two modules: group construction by block matching and SVD-based filtering. The collaborative filtering is has been achieved determined by an iterative singular value decomposition (SVD) that calculates the singular values with their singular vectors by an iterative power method whose with an internal data type utilizes using double precision floating-point representation. To adapt the video coding standard as well as being to be hardware friendly, this proposal addresses this issue by eliminating the double precision values via the decimal digits clipping after shifting the intermediate results to large numbers during iterations. The simulation results reportedly show that the proposed fixed-point algorithm for the SVD module could can achieve comparable performance with the original NLSF algorithm.

The bBit rate reduction compared to JEM7 (with all tools on) is reportedly 0.86% for RA CfP conditions, whereas the FP floating-point implementation of JVET-J0011 reportedly gave 1.25%. Similarly for LDB, 1.64% was reported for the integer method, and 1.92% FP for the floating-point version.

21 groups are used, with patches are of size 6x6. A total of 21 SVD analyses has to beare determinedconducted, which could consist of 36 basis functions at a maximum, but it is reported that due to thresholding only 4.6 basis functions on average need to be computed. A maximum of 10 was found necessary.

The dDecoder runtime is reportedly increased by 316% (compared to 397% in FPwith the floating-point scheme).

The gGrouping/clustering, determination of the SVD basis, and computation of the SVD based reconstruction is necessary at both the encoder and decoder side.

Question asked in the discussion were:

• 
  It was askd Hhow many operations are needed for the grouping.? (This could be the main reason for the complexity.)
  
• 
  It was asked hHow many operations are needed to determine one SVD basis at a maximum (e.g., when restricting the scheme to 10 basis functions).?
  

The fFiltering itself is probably is of less concern than the SVD aspects.

JVET-J0077 Deblocking Improvements for Large CUs [W. Zhu, K. Misra, A. Segall (Sharp)]

The contribution proposes a deblocking process designed to reduce the blocking arteifacts that result from the use of large transforms and block sizes. Compared to the JEM deblocking approach, the process incorporates stronger filters for both luma and chroma. Additionally, the process includes a control process that considers the block sizes on both sides of the boundary being deblocked. The stronger filters are used for luma samples that correspond to larger block sizes, while the JEM deblocking filters are still used for luma samples corresponding to the smaller block sizes. From chroma, the filter is selected usinges a different approach, and the stronger filter is applied when chroma samples on either sides of the deblocking boundary belong to a large block. It is reported that the proposed deblocking change improves subjective quality at low bit -rates when compared with the deblocking used in JEM7, and it is proposed to include the technique in formal study.

Revision 1 of the contribution includes a modification to the deblocking control process resulting that results in application of wider and stronger filter for large block boundaries when blocks on either side of the boundary make use of lLocal Iillumination cCompensation and the CBF is 0 for that block.
It wais suggested to deblock 7 samples for large blocks and 3 samples for small blocks.

The results shown with still picture snapshots weare obvious, and proponents believe that it is also visible in video. More extensive viewing tests would be necessary to assess the subjective effect on video.

For further studyFurther study of this was requested.