Adaptive MMVD offset selection improves video decoding efficiency by balancing motion vector precision and prediction accuracy.
Inverse quantization and up-sampling decompress image data without entropy decoding, de-zigzag, or inverse transform hardware.
Pre-decoded region search limits intra block copy reference blocks, improving video compression efficiency while controlling complexity.
Different QPs for basic and additional views enable merged multi-view bitstreams that cut bandwidth use while limiting quality loss.
Limits intra block copy references to pre-decoded regions, improving video compression efficiency while reducing memory and processing load.
Constraining intra block copy search regions improves video compression and image quality while limiting encoding complexity and memory use.
Neighboring MIP block data sets default intra prediction modes, improving video compression efficiency while preserving picture quality.
Adaptive QP assignment by feature channel improves AI-oriented compression efficiency while preserving critical task-relevant information.