Efficient algorithm and architecture designs for MPEG-4 shape adaptive video object coding

Abstract

An efficient algorithm and its VLSI architecture design for real-time shape adaptive zerotree coding (SA-ZTC) are presented. The proposed algorithm takes the DWT coefficients of the image as input and encodes the quantized coefficients efficiently to achieve the data compression. Most other wavelet coefficient-coding algorithms cannot provide real-time computation because they use multi-pass algorithms to decide the zerotrees formed by the wavelet coefficients. In contrast, the proposed SA-ZTC adopts a one pass algorithm and thus greatly reduces the computation time for real-time processing. Besides, the proposed architecture can be equally applied to both whole image coding without shape mask (eg, JPEG 2000) and object-based image coding with shape mask (eg, MPEG-4). Experimental results indicate the proposed algorithm can achieve better coding efficiency than the widely adopted EZW and the performance is close to the multi-pass SPHIT algorithm. The derived architecture design is shown to be hardware cost-effective and capable of delivering real-time processing.