Abstract: Edge extraction plays an improtant role in automatic processing of SAR images. It is difficult for conventional edge extraction methods to take account of noise suppression, the integrity of edges and the veracity of position synchronously due to the existence of the stronger speckle. Based on the analysis of the characteristics of SAR images, a fuzzy fusion algorithm for edge extraction is presented in this paper, which is composed of multi-direction, multi-scale wavelet transform and fuzzy integral in terms of the relationship between mutli-scale edge grads information and the mutual supplement of multi-direction edge grads information. Experimental results show that the proposed algorithm can suppress noise and preserve weak edges effectively, and that the fused edges have precise positions and an intact contour.

Key words: multi-direction wavelet transform, fuzzy integral, edges fusion, SAR images