SAR IMAGE DESPECKLING WITH THE MULTI-SCALE NONLOCAL LOW-RANK MODEL

Motivated by the idea of low-rank prior, we propose a despeckling method based on multi-scale nonlocal low-rank model. Specially, the proposed low-rank model consists of a data fidelity term derived from the Fisher-Tippett logarithmic- space speckle distribution and a weighted nuclear norm regularization term. Furthermore, we exploit a multi-scale prior by selecting similar patches from different scales of the SAR image. The resulting optimization problem is solved by the alternating direction method of multipliers (ADMM). Experiments conducted on one real SAR image demonstrate that the proposed method can achieve comparable and even better despeckling results than state-of-the-art SAR despeckling methods, both visually and quantitatively.