Image Compression using Adaptive Wavelet Packet and Multistage Vector Quantization

Abstract

This paper presents a new image coding technique using adaptive wavelet packet and multistage vector quantization. Wavelet packets are generalization of wavelet transform, capable of providing arbitrary frequency resolution to meet signal's spectral behavior. Image properties, filter and cost function are the three prime factors which are commonly used to select wavelet packet basis. In this paper, the best basis is selected through singular value decomposition. After selecting the best tree, the coefficients of the best tree are quantized using multistage vector quantization. Experimental results show that wavelet packet transform brings consistent improvement over dyadic wavelet transform.