Image Coding by Usning Smooth Side-Match Classified Vector Quantizer with Variable Block Size

Abstract

Although the side-math vector quantizer (SMVQ) is efficient in reducing the bit rate, the image coding quality is in general degenerates. The reason is that if the gray level transition across the boundaries between the neighboring blocks is increasing or decreasing, SMVQ may not encode the block well. In this study, we proposed a smooth side-match method to select the state codebook according to the smoothness of the gray levels between the neighboring blocks. This method achieves the higher PSNR and the better visual perception quality than SMVQ does while the bit rate is the same. Moreover, a genetic clustering algorithm that can automatically find the proper number clusters is proposed to design the codebooks because the well-know algorithm LBG has the shortcoming of requiring the user to supply it with the number of clusters. Therefore, the proposed smooth side-match classified vector quantizer (SSM-CVQ) is the combination of three techniques: classified vector quantization, variable block size segmentation and the smooth side-match method. As indicated by the experimental results, SSM-CVQ has the higher coding quality and the lower bit rate than other methods have. The Lena image can be coded by SSM-CVQ with 0.172 bpp and 32.49 dB.