KSR-NLM: a non local means despeckling filter for ultrasound images based on ratio patch and KS distance.

Speckle noise greatly degrades the quality of ultrasound images. Being signal dependent, it requires the design of specific filters in order to be reduce. Within this manuscript, $a$ novel approach for despeckling ultrasound images is proposed. The methodology belongs to the Non Local Means family. The novelty consists in the methodology adopted for measuring patches similarity. In brief, the statistical distribution of the ratio image patch is estimated and compared to the theoretical Cumulative Distribution Function. More in detail, the Kolmogorov-Smirnov distance is adopted for measuring the similarity between the two distribution. The method, namely KSR-NLM, has shown to achieve good denoising performances both in case of synthetic and real datasets.