Coefficient-Tracking Speckle Reducing Anisotropic Diffusion

Speckle reducing anisotropic diffusion (SRAD) filter is introduced to significantly reduce speckle noise from images. Yet, SRAD suffers from the problems of ordinary diffusion filters, e.g., objects boundaries broadening and edges dislocation. This paper provides a more robust diffusion-filtering scheme, which is based on tracking the image main features across SRAD wale-space images. Coefficient-tracking SRAD (CSRAD) controls the amount of allowed diffusion based on the edges original location. CSRAD is tested on Berkley segmentation dataset. CSRAD results are Subjectively compared with those of SRAD in terms of edge localization, smoothing enhancement, and features preserving. Experimental results show that CSRAD significantly reduced the features distortion and edges dislocation effects. Consequently, the entire diffusion process is enhanced.