Embedded wavelet-based coding of the shape and texture of arbitrarily shaped visual objects.

摘要

This thesis addresses the issue of coding the shape and texture of arbitrarily-shaped visual objects. Arbitrarily-shaped visual objects are images which are not necessarily rectangular in shape and are described by two basic features: shape and texture. Current state-of-the-art object-based coding schemes do not provide integrated coding of both the shape and the texture. They require that the shape of the object be pre-transmitted before the texture can be decoded.; An embedded wavelet-based scheme which codes both the shape and texture of an arbitrarily-shaped object in one integrated algorithm is proposed. Called shape and texture SPIHT (ST-SPIHT), the proposed scheme is based on the shape-adaptive discrete wavelet transform (SA-DWT) with lifting implementation and the set partitioning in hierarchical trees (SPIHT) wavelet-based coding scheme. Subjective and objective results showed that the ST-SPIHT scheme performed better than other state-of-the-art schemes, such as JPEG2000, under certain conditions and requirements.