Abstract: In order to reduce speckle noise in SAR image processing while preserving scatter targets and edge as more as possible, an adaptive speckle reduction method based on a kernel regression is presented. By analyzing the magnitude distribution characteristic of SAR image, while building model the image magnitude is chosen as the classification condition. The kernel function heavily smoothes to reduce the speckle for background region with small magnitude, and protect targets for targets region with large magnitude. Then considering preserving the edges, the steering kernel is modified based on scatter matrix, finally the speckle reduction method based on kernel regression for SAR image is proposed. The experiment results show that the proposed method can reduce speckle noise while preserving targets and edges by introducing magnitude information and scatter matrix into kernel function.

Key words: speckle, SAR image, kernel regression, adaptive, magnitude