Research For Codebook Of VQ And Achievement Of Correspond Decoding Circuit

Vector Quantization(VQ), is used largely in design for ASIC for image compress because of its simple algorithm and easy realization in hardware. Image database for character and algorithm for searching are keys of VQ. VQ technology today we used is expensive when mapping to hardware and much delay in time, so it is not fit for moving pictures. This paper resolve the problem by new VQ algorithm, it can improve rate for encoding in effect of ASIC for image compress . The corresponding circuit for decoding is simple to realize. Through experiments, it shows that quality of rebuilt image is better and encoding rate increase largely. This makes it possible to transmit pictures in general communication nets.The paper utilizes advantage of wavelet in processing digital signals and ocular character of image , analyzes image vector through wavelet, distributes codebook four part as vertical, horizontal, diagonally and same, in each part, we get characteristic database, so that we distribute traditional database to four directional databases. When encode, we confirm direction of image vector first, then, match in corresponding database. By this way, time for encode can be reduced largely, but PSNR not. Experiment shows that average ratio for encode can increase 38. 4%, at most 45. 8%, butPSNR only reduce 1. 8%.Secondly, the paper sets up architecture for hardware based on directional encoding algorithm. Also, it predigests judgment rules to reduce circuits scale. Gate-level simulation shows that it is right and rational.Thirdly, the paper achieves circuit structure for decode correspond to PDVQ(Partition Dynamically VQ), RTL codes and verifies its function. Accordance of result for gate-level testing and software simulation shows that the design is correct.At last, as a whole VQ system, the paper also uses RTL code to describe VQ system by new arithmetic and verify it right.