Adaptive geometric wavelet transform based two dimensional data compression

Abstract

Easy Path Wavelet Transform (EPWT), a wavelet based compression technique that averages in adaptive neighborhoods of image intensity points, is utilized for the sparse representation of two-dimensional data. EPWT, when applied upon images, operates along the permutations of original image indices exploiting the local correlations in an uncomplicated manner. The highly correlated array of function values when passed through one-dimensional wavelet filters produce wavelet coefficients of minimum magnitude. Consequently high compression ratio is achieved after the application of a wavelet shrinkage procedure on the reconstructed images. Thus the image can be represented in terms of its wavelet decomposition coefficients and the respective path vectors obtained at all levels.