Systematic codebook design method in Adaptive Resolution Vector Quantization technique for compound image compression

Abstract

Conventional image compression techniques based on transform coding to frequency domain have a problem of image quality, when images contain sharp edges like character patterns are compresed. In order to solve this problem, we have developed Adaptive Resolution Vector Quantization (AR-VQ) method and a systematic codebook design method applied to all kinds of images without using learning sequences for 4x4 and 2x2 pixel blocks. By using these methods, consequently, we can realize much superior compression performance than the JPEG and the JPEG2000. On the compression of the XGA (1024x768 pixels) images including text, for instance, there exist an overwhelming performance difference of 5 to 40 dB in compressed image quality.