Image compression using linear prediction operator of wavelet coefficients

In this paper, we propose a new prediction operator called the linear prediction operator. The linear prediction operator is based on the explicit model of statistical relationship of the high-frequency coefficients between the different frequency sub-bands and the scaled similarity of the high-frequency coefficients with respect to their orientations. This scaled similarity is exploited to approximate the functions in different resolutions to achieve high compression ratio.

Keywords: Fractal coding, haar transform, multi resolution analysis, nonlinear approximation, zero-tree wavelet

doi :10.4038/jnsfsr.v37i3.1210

J.Natn.Sci.Foundation Sri Lanka 2009 37 (3): 167-174