Extended Transform Selection in Video Coding

The transform selection process is segmented into two stages: first selecting between DST and DCT based on prediction mode, then selecting specific transform types based on block size. This segmentation reduces the overall complexity by breaking down the transform selection into manageable decision trees rather than evaluating all transform options simultaneously.

The transform selection is made dynamic and adaptive based on coding conditions. The encoder dynamically selects transforms from the extended set based on prediction mode and block size, allowing the system to adapt to different coding scenarios while maintaining manageable complexity through conditional selection rules.

Different transforms are assigned to different local coding conditions (prediction modes and block sizes). Specifically, DST transforms are used for certain prediction modes while DCT transforms are used for others, and specific transform types are selected based on block dimensions. This local optimization allows the system to achieve better compression performance by matching transforms to local characteristics without requiring all transforms to be evaluated everywhere.

Instead of using all available transforms for all blocks, the system applies transforms selectively based on coding conditions. The extended transform set provides more options, but only the appropriate transforms are applied to each block based on prediction mode and size, avoiding unnecessary computation while maintaining the ability to achieve superior compression when needed.