Joint Collaborative Team on Video Coding (jct-vc)

5.10Motion compensation operation and interpolation filters

5.10.1Interpolation filters and MV precision

JCTVC-I0550 Sum-128 luma/chroma interpolation filter [H. Zhu (personal participant)] [late]

A luma/chroma interpolation filter for which the sum of all coefficients are 128 is proposed in the document. Simulation were reportedly conducted under the common test conditions using the reference software HM5.0 r1682. The results were reportedly −0.1%Y, 0.1%U, 0.1%V for RA-HE, 0.1%Y, 0.1%U, 0.2%V for RA-Main, 0.0%Y, 0.1%U, 0.2%V for LB-HE, 0.5%Y, −0.2%U, −0.4%V for LB-Main, −0.2%Y, 0.1%U, −0.2%V for LP-HE and −1.8%Y, −1.4%U, −1.6%V for LP-Main.

The document was not uploaded until Sun 04/29. The presenter was not available to present it on the several occasions during the meeting at which an opportunity was available for its presentation. No action was taken on this.

5.10.2Weighted prediction

JCTVC-I0109 New results of implicit weighted prediction [A. Tanizawa, T. Chujoh, T. Yamakage (Toshiba)]

This document reports results for implicit weighted prediction on the IBBBP (M=4) coding structure. An AVC-based weighted prediction scheme has been adopted in the HM and WD. Explicit WP can reportedly achieve large coding gains for fading sequences with illumination variation. However, the gain of the implicit WP compared to the explicit WP on current coding structures (RA and LDB) is reportedly not so high. The main reason is reportedly that the AVC based implicit WP design was focused on the IBBP (M=3) coding structure originally and is not as effective in the current coding structures of the HM common test conditions. For example, in the case of the RA CTC coding structure, some weighting factors deriving from the implicit WP are identical to the default weighting factors. In this contribution, it is reported that the implicit WP on IBBBP (M=4) coding structure is tested and reported. It was reported that AVC-based implicit WP can get 3.3% BD-BR gain on average for black-fade/white-fade sequences.

Implicit WP was only used for B pictures.

Explicit WP gives roughly 20% BR gain for the same fade sequences.

It was asked whether implicit WP is useful at all. Even the one in AVC seems to be hardly used, and it seemingly cannot compete with the explicit signalling.

JCTVC-I0110 Improved implicit weighted prediction [A. Tanizawa, T. Chujoh, T. Yamakage (Toshiba)]

This document presents an implicit weighted prediction design. The weighted prediction (WP) scheme consists of two modes, an explicit WP and an implicit WP. An implicit WP is a method of not explicitly encoding WP parameters, but implicitly deriving them from information of reference pictures. In AVC-based implicit WP, the weighting factors can be derived from the temporal distances among the target picture and two reference pictures are used for bi-prediction in B-slices. However, the offset value is always set to 0. It was reported that there are issues in that the conventional implicit WP cannot compensate the pixel value additively and cannot be applicable to uni-prediction. A derivation process was proposed for implicit WP parameters based on image characteristics of the reference pictures stored in decoded picture buffer. The experimental results in HM software version 6.0 including AVC-based implicit WP are reported. The BD-BR gain of the proposed method for black/white-fade sequences is reportedly 16% to 34% compared to the HM common condition anchor. It is reported that this scheme can reduce the decoding time.

The reduction of decoding time is likely due to have a lower bit rate (less CABAC processing)

Here, the identical processing for deriving the WP weights is simply shifted to the decoder. Even then, it performs slightly worse than the explicit case with the identical algorithm. Several experts expressed the opinion that this is not desirable (more decoding complexity, error resilience issues, too specific for fades, and explicit WP would be needed anyway).

JCTVC-I0128 Cross-check report of Toshiba proposal JCTVC-I0110 [P. Bordes (Technicolor)] [late]
JCTVC-I0115 Improvement of Implicit Weighted Prediction [P. Bordes, P. Andrivon (Technicolor)]

This document presents an implicit weighted prediction (IWP) design. Weighted prediction (WP) consists of two modes: an explicit WP mode and an implicit WP mode. In explicit WP, the weighting parameters are transmitted explicitly, while in implicit WP they are derived from the temporal distances of the two reference pictures used for bi-prediction in B-slices.

In the current IWP, the offsets are unforced to zero and it is reportedly applicable to bi-prediction and B-slices only. In RA (hierarchical GOP) the weights of the closest references are equal to defaults.

The proposed IWP modifies with these limitations by interpolating or extrapolating the explicit WP parameters, in a case where explicit and implicit WP are combined.

The experimental results in HM software version 6.0 are reported. The BD-BR gain depends on the proportion of frames using explicit WP and implicit WP (EP=Explicit Period). For EP=4, the BD-BR gain of the proposed method for black/white-fade sequences is reportedly 17% to 34%.

Implicit weights are interpolated from explicit weights (e.g. the two reference pictures in a B frame pyramid).

This could rather be interpreted as a method of coding for the explicit parameters, as it cannot replace the explicit signalling. Also here it does not give a benefit in all cases compared to explicit signalling, and the maximum benefit would be around 0.2–0.3%.
JCTVC-I0111 Cross-check of implicit weighted prediction (JCTVC-I0115) [A. Tanizawa, T. Chujoh (Toshiba)]
JCTVC-I0589 Text for the removal of implicit weighted prediction [T. Hellman (Broadcom)] [late]

Conclusion on I0109, I0110 and I0115: explicit WP is still the most universal and best choice in terms of compression. The current implicit WP design is useless. Decision: Remove implicit WP from CD. (T. Hellman volunteered to provide the text (new doc) and software.)

Decision: Adopt.