Orientation information measure based image restoration
Xinnian Wang, Dequn Liang · 2004
Various psychophysical experiments have shown that the visibility of noise is greatly masked by sharp intensity transitions, whereas blurring generally appears to be unacceptable in this context. According to this property of human visual system, in this paper, we propose an image restoration algorithm based on orientation information measure. By this method, the amount of restoration in the flat image regions is small so that the noise will not be magnified to have dominant effect, whereas in the regions with sharp intensity transitions, the amount of restoration should be large so that sharp edges can be reconstructed. It is shown by the experiments that the restored images obtained by the proposed algorithm are better in terms of visual quality.