Image coding technique using M band filter bank and classified vector quantizer

In this paper, we propose an efficient classified vector quantizer (CVQ) technique employing the linear phase para-unitary (LPPU) M band filter bank. Based on the asymptotic distortion of VQ, we derive a new coding gain. Then, we design the LPPU filter bank which maximizes the coding gain. Employing the LPPU 4 band filter bank, we propose a CVQ technique. We can classify the input image more accurately by decomposing the input image into 4×4 blocks by LPPU 4 band filter bank, compared to the case of 8×8 blocks. And we propose a technique to reduce the side information for the classification. Simulation results indicate that the proposed scheme yields about 0.8~1.0 dB PSNR gain over the JPEG (Joint Photographic Experts Group) in the range of 0.3~0.4 bits per pixel (bpp), for Lena image