Isointensity directional smoothing for edge-preserving noise reduction

Isointensity directional smoothing is a novel edge-preserving smoothing method. By fitting a polynomial hypersurface to the local intensity, isointensity bands (isobands) are derived which approximately parallel an edge when the centre pixel is close to the edge. Pixels in the isoband passing through the centre are assigned higher weights, while those in other isobands are assigned lower weights according to a decaying function of the estimated intensity distance between the pixel and the centre. The weighted average of the pixel values is used as the smoothed value at the centre. The proposed method has a strong smoothing effect while causes little blurring of the edge, and the preserved edge has a good formation. The filter also performs well in intensity slope area. Experiments were carried on synthetic and real images. The results showed that the proposed filter has better performance than some well known edge-preserving filters compared in this work.