High-Efficiency Weighted Prediction for H.265/HEVC Standard

To achieve high coding efficiency, hybrid video coding consisting of block-based motion compensation (MC) and block-based transform is widely used in many video coding standards such as MPEG-4 and H.264 Advanced Video Coding (AVC). Since temporally successive video images are highly correlated, redundant data in successive frames can be reduced by using MC. However, MC does not work well when temporal illumination variations such as fading and dissolving are present because MC relies on motion estimation based on a block matching technique.

To resolve this issue, Toshiba has developed a weighted prediction (WP) technology with a view to its adoption as part of the H.265 High Efficiency Video Coding (HEVC) standard. H.265/HEVC dramatically improves coding efficiency for such images by performing linear pixel value prediction using a multiplicative weighting factor and an additive offset after MC. In order to determine the optimal WP parameters for an encoder, we have proposed a WP parameter derivation method using image characteristics between the target and reference pictures.

This method is incorporated into the H.265/HEVC reference software. H.265/HEVC using the WP technology provides much higher coding efficiency than H.264/AVC and is expected to realize the distribution of high-quality video contents to various audiovisual devices including tablets, TV sets, and PCs.

MPEG-4: Moving Picture Experts Group Phase 4

Outline of motion-compensated WP for H.265/HEVC standard