SAR image despeckling via modified non-local means: based on SURE criterion

Aimed at the shortage of similar region capture and directional information obtainment for SAR image despeckling using conventional non-local means method (NLM), a new NLM SAR image despeckling method is proposed based on multiple different directional anisotropic Gaussian directional window and Stein unbiased risk estimation (SURE) aggregation. The ratio measurement strategy is utilized to compute the similarity of two patches and the NLM result is computed based on the anisotropic Gaussian windows with some direction. The results of NLM with different anisotropic Gaussian windows are aggregated by using the Stein unbiased risk estimation criterion to obtain the final SAR despeckling result. For multiple SAR images, the experiment results show that the new method has advantages in the SAR image despeckling performance, and can well preserve the local geometric structure information, which is essential for understanding and interpretation of SAR image.