A Novel Strong Decorrelation Approach for Image Subband Coding Using Polynomial EVD Algorithms

Abstract

Subband coding is a popular technique to achieve multichannel data compression and efficient data transmission in image and video communications. In this paper, we focus on designing a data-dependent image subband coder which divides the image into strongly (total) decorrelated and spectrally majorized subbands, which seeks to reduce the redundancy to achieve better compression and efficient transmission. To achieve strong decorrelation and spectral majorization between the subbands, we adopt a set of new iterative polynomial eigenvalue decomposition (PEVD) algorithms: sequential matrix diagonalization (SMD) and maximum element sequential matrix diagonalization (ME-SMD). Using this SMD-based PEVD approach, we design the data dependent subband coder (DDSC) for image subband coding. We compare the performance of the proposed SMD/ME-SMD algorithms with the existing DDSC methods like SBR2, SBR2C, K–L transform (KLT) coder and data independent subband coder (DISC)-based discrete wavelet transform (DWT) technique. To measure the performance, we use the parameters like coding gain, correlation coefficient, MSE (mean square error) and peak signal-to-noise ratio (PSNR). The presented simulation results for standard images in the absence of quantization show that the proposed SMD-based PEVD technique performs far better than the existing techniques.