DESIGN TREES FOR ADAPTIVE CODING OF IMAGES AND VIDEOS USING SET PARTITIONING IN GENERALIZED HIERARCHICAL TREES HAVING DIRECTIONALITY

摘要

Adaptive entropy encoding and decoding which utilizes Set Partitioning within Generalized Hierarchical Trees (SPRIGHT) and a method of designing trees utilizing directionality. After decorrelation and quantization a tree structure is selected from multiple candidates, based on geometric relationships within the image block, for coding the coefficients toward improving zero-clustering of coefficients. Trees for the SPRIGHT encoding are created in response to finding frequency position of each coefficient and scaling frequency position followed by use of octave-band partitioning of coefficient patterns into squares and L-shapes, and the L-shapes are iteratively partitioned into squares. The tree comprises leaf nodes containing coefficients associated with each non-leaf node. The number of zero clustered coefficients can be increased, thus decreasing the number of nodes coded into the encoded image output.