Subband decomposition based image compression algorithms with nonlinear adaptive filter banks.
Ruşen Öktem, Karen Egiazarian, A. Enis Çetin · 1999
Image compression via adaptive perfect reconstruction filter banks is studied. An M-channel filter bank is employed in a lifting scheme using MSE-adaptive predictors. FFT-ordered L-filters are used for prediction. L-filters include a class of linear and nonlinear filters. A switching between three different filters is made for applying different filters over smooth regions, around edges and over texture regions of the image. FFT-ordering enables the use of different filters within the same implementation structure by only changing the node operations. Experimental results show that switchingimproves the performance in both subjective and objective terms in all regions. 1. INTRODUCTION Subband decomposition has been widely used in many signal processing applications includingimage compression. In most of the image compression applications, subband decomposition is achieved by a pair of linear time invariant filters. However, linear time invariant filters suffer from ringingand aliasing...