SAR image despeckling based on improved non-local means algorithm

To efficiently preserve tiny details and sharpness information of synthetic aperture radar (SAR) images while clearly remove the speckles, an improved non-local means method is proposed. Firstly, aiming at the traditional patch-similarity measurement not precise using in SAR image, we define a new variable to obtain a more precise patch-similarity measurement results. Furthermore, the PDF mapping function using for calculating ratio distance is also modified to get the patch-similarity between pixels more accurate. The experimental results show that both the visual quality and evaluation index of the proposed method outperform the traditional three methods: enhance Lee filtering (ELF), the bilateral Filtering (BF) and the NL-means method.