An Efficient Fuzzy Possibilistic code book design for Vector Quantization based image compression in the Wavelet Packet Domain

This paper presents an efficient fuzzy possibilistic code book design for vector quantization in the wavelet packet domain. Wavelet packet tree (WPT) methodology is applied to the whole image. The sub blocks of the input image are decomposed into two level WPT where all coefficients of LL band and the approximation coefficients of LH, HL & HH bands are quantized using the proposed vector quantizer. The quantized coefficients are further compressed using Huffman encoder and then transmitted across. The image is reconstructed using the inverse WPT followed by index reassignment and the subsequent decoding process. The efficiency of the proposed work is analyzed by varying the code vector (cluster) size from 8 to 512 in the order of 2ⁿ. It is found that the code vector of size 8 × 8 stands as a good choice by maintaining a compromise between quality and compression. The proposed work is also compared with other existing techniques. The results show that the psycho-visual fidelity criteria (both subjective and objective measures) of the proposed work are better than the other existing techniques.