A Digital Image Compression Technique Based on Dynamic State Codebook

• Main Authors: Cheng-Ta Huang, 黃程達
• Other Authors: Wen-Kuang Chou
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/5q5tj9

碩士 === 靜宜大學 === 資訊管理學系研究所 === 94 === Vector Quantization (VQ) is a popular digital image compression technique. VQ can compress digital images with a very small capacity of storage space in a fast, simple and efficient process. It is an important technique with fast compress and small size of storage in current of information and network environment. Many new schemes based on VQ have been proposed. In this thesis, we propose a new scheme that can further compress VQ index table to achieve better compression rate. To speed up our process, we sort VQ codebook with Principal Component Analysis (PCA). By using PCA, we can make the neighboring codewords with maximum similarity and correlation. And then we apply Search-Order Coding (SOC) to find whether the neighboring blocks are the same as the encoding block. If SOC fails to encode the current block, we then apply Side Match Vector Quantization (SMVQ) with dynamic state codebook to encode it. To increase the performance of SMVQ, we classify the current block into ten groups with different state codebook size by referring the variances of its upper and left blocks. The experiment results confirm our proposed scheme provides lower compression ratios without distortion than Chang et al.’s and Chen et al.’s schemes did.