Hybrid Vector Quantization Design Using SPIHT and PVQ in the Lifting Wavelet Domain

Abstract

In this paper, an efficient vector quantization scheme which allows easy expansion of the original vector quantizer design to operate at higher bit rates for lower distortion is presented. Image compression based on wavelets results in high compression ratio compared to other compression techniques. In the wavele decomposition of the image, the lowest frequency sub band contains most of the information. The existing literature consists of compression algorithm in which multi stage vector quantization is applied to lowest frequency sub band and pyramidal vector quantization to high frequency sub band. To have better PSNR values, we have applied SPIHT to low frequency coefficients and PVQ to high frequency co efficients. Experimental results show that the combination of SPIHT and PVQ outperforms the combination of MSVQ with PVQ in terms of PSNR. Our implementation offers exact reconstruction, low computational complexity and ease of implementation. The proposed method shows better image quality in terms of PSNR at the same compression ratio as compared to other DWT and VQ based image compression techniques found in the literature. The proposed method of image compression is useful for various applications in which high quality (i.e. high precision) is crucial. The experimental results reveal the effectiveness of the proposed hybrid scheme in image compression.