Image compression scheme using improved basic-lavq and optimized VLC

Abstract

In this paper, a predictive locally adaptive vector quantization (PLA VQ) for real-time image compression is developed and simulated. It is a hybrid of improved predictor, LA VQ and (VLC). The lat-ter is a combination of Laplacian quantizer, Huffman coding, and optimum-modified Bl-code for encod-ing the codewords and indices, respectively. Simulation shows that an improvement in speed of about 76% can be achieved with about 1.1 dB increase in the peak signal-to-noise ratio (PSNR) at the same bit rate. Optimization of the modified B I-code decreases the bit rate with an average of 37% for the same image quality. Simulation data of PSNR and bit rate vs. error threshold can be expressed by a decaying exponen-tial model independent of image type.