Nonlinear weighted median filters in dyadic decomposition of images

Abstract

There are many techniques of loss image compression based on dyadic decomposition of images. In these decompositions always linear filters are used. In this paper it is proposed to use nonlinear weighted median filters in the place of linear ones and it is shown how median filters may be used in dyadic decomposition. This technique has been compared with the non-standard wavelet decomposition, in particular with the Haar wavelet filtering. The experiments made show that in the loss image compression case using median filters gave better visual quality of the reconstructed images.