Comopact image representation using wavelet lifting along arbitrary trees

Authors:

Abstract
In this work, we present a method for achieving a compact image representation by applying wavelet lifting along an arbitrary combination of trees. This is done by extending a lifting transform along a single tree to a series of lifting transforms along a number of different trees, each of which exploits directionality differently. The trees are constructed in ways that force tree edges to follow the geometric flows inherent in images. In particular, we propose a method for constructing trees that trace out geometric flows using edges in the image. Our preliminary results demonstrate promising performance in terms of PSNR as a function of the percentage of retained coefficients, motivating its potential for image coding.