A Method of SAR Image Speckle Reduction with Feature Preservation Algorithm in Frequency Domain

On the basis of the speckle reduction and feature preservation of SAR image, a new algorithm based on multi-scale auto-convolution (MSA) and phase congruency (PC) is proposed for SAR image speckle reduction. The multi-scale average property of MSA and the characteristic of good localization of feature of PC in Fourier transform domain are combined together. At first, different filtered regions are analyzed and judged in SAR image by the proposed algorithm. Then, the corresponding filtering strategy is adopted. The MSA filtering method is implemented in proportional and homogeneous regions. The enhanced MSA filtering method by measure of phase congruency is performed in the analyzed and judged as feature regions according to the measure of phase congruency. The average MSA filtering method is used in the isolated point object regions. The conflict of spatial adaptive filtering methods between speckle reduction and feature preservation is deal with in SAR image speckle noise reduction. MiniSAR image data is used for verification and the experiment results indicate that the proposed algorithm has a good performance of speckle noise reduction.