Sung-Moon Chun

Shape-adaptive region partitioning method for shape-assisted block-based texture coding

In the content-based image coding scheme, segmentation information of the arbitrarily shaped regions may be available for both encoder and decoder. The shape-assisted block-based texture coding methodologies, such as shape-adaptive discrete cosine transform (SADCT), can use this segmentation information to improve the coding efficiency. We introduce the shape-adaptive region partitioning (SARP) methods which can reduce the number of coding blocks that partition the arbitrarily shaped region by modifying the block positions. By simply adding the SARP method to aid the SADCT, the coded texture bits can be reduced by 5-10%, in comparison with the SADCT using a common block-based coding infrastructure which is usual for MPEG-1/2 and H.263.