Selective Reference Line Grouping for Video Coding

The patent segments the reference lines into multiple groups (first reference line group, second reference line group, etc.) based on their distances from the current block. This segmentation allows the decoder to selectively process only relevant reference line groups, reducing the computational complexity while maintaining prediction accuracy by using the appropriate number of reference lines for different block sizes and positions.

The patent applies partial action by selectively using only the necessary number of reference line groups based on the current block's characteristics (size, position). Instead of always processing all available reference lines, the method uses a flag (mrl_flag) to indicate whether to use one or multiple reference line groups, thereby reducing unnecessary computational operations while maintaining prediction accuracy where needed.

The patent introduces dynamic adaptability by using the mrl_flag to dynamically select between using one reference line group or multiple reference line groups based on the current block's characteristics. This dynamic approach allows the encoder to adjust the number of reference lines used for prediction, optimizing video quality while controlling encoding complexity on a per-block basis rather than using a fixed approach for all blocks.

The patent applies local quality by treating different blocks differently based on their specific characteristics (size, position, content). Some blocks may benefit from multiple reference line groups while others may not, and the mrl_flag enables this localized optimization. This ensures that video quality is improved where multiple reference lines provide benefit, while avoiding unnecessary complexity in blocks where a single reference line group is sufficient.

The patent applies partial action by using only the necessary number of reference line groups (one or multiple) based on the mrl_flag and block characteristics. This selective approach reduces the average processing time across all blocks by avoiding the use of multiple reference line groups when they are not needed, while still achieving good prediction performance for blocks that benefit from multiple reference lines.

The patent performs preliminary action by determining the mrl_flag and selecting which reference line groups to use before actually performing the prediction operation. This preliminary selection based on block characteristics (size, position) and the mrl_flag avoids unnecessary processing of multiple reference line groups for blocks that would be adequately predicted using a single reference line group, thereby reducing overall processing time.