Enhancement of wavelet-coded images via novel directional filtering

Abstract

In many multimedia applications image compression is required to substantially reduce the amount of image data. This compression, however, sometimes brings artifacts. Typical artifacts are blocking artifacts and mosquito noise in DCT-coded images, and ringing artifacts around edges in wavelet-coded images. We propose a new directional post-processing algorithm which can reduce ringing artifacts around edges of a compressed image. The algorithm detects edge direction in each N/spl times/N block and then applies an appropriate window shape to decoded N/spl times/N block images for nonlinear filtering. Simulation results show that the proposed algorithm is as effective as or more effective than existing nonlinear filtering techniques.