Codebook Generation for Vector Quantization with Edge Features

In this paper, we propose a new technique to retain the edges of the images while compressing them. The image is divided into small blocks of size 4 x 4 pixels. From the list of N image blocks, M (M < N and M is the desired size of codebook) representative blocks are selected to form the codebook. In ordinary codebook generation, all the image blocks are treated equally. But, in the proposed method, the image blocks are classified into high detail blocks and low detail blocks. The high detail blocks represent the edges and the low detail blocks represent the shaded blocks. The high detail blocks are given preference while generating the codebook. Categorization of image blocks improves the quality of the reconstructed images. Codebooks of sizes (M) 128 and 256 are created with the images of size 256 x 256 pixels. The proposed method gives better performance when compared with the results of few existing techniques.