An Improved Multi-stage Vector Quantization for Image Coding

The traditional full search vector quantization (FSVQ) can give a desirable quality for image coding, but it consumes too much time and memory space. The multistage vector quantization (MSVQ) was proposed to alleviate the computational complexity of FSVQ. However, MSVQ cannot give a comparable performance as FSVQ for image coding. To better tradeoff the coding performance and computational complexity, we propose a classified multistage vector quantization (CMSVQ) in this paper. The simulation results show that the CMSVQ scheme exhibits higher codewords´ use efficiency a better trade-off between the image quality and the computational complexity.