Ultrasound despeckling based on Non Local Means

Abstract

Ultrasound images are characterized by speckle, a multiplicative noise that degrades their quality. In the last decades, several efforts have been done for developing effective denoising filters able to provide effective signal regularization and noise preservation. Recently, the so-called Non Local Mean approaches have proven to be well suited for such kind of problems. Within this manuscript, a new despeckling filter for ultrasound data is presented, developed in the Non Local Mean framework that jointly exploits several acquired video frames for reducing speckle. The main novelty consists in the metric adopted for the evaluation of patches similarity, which is based on the statistical properties of the acquired data. More in detail, the Kolmogorov-Smirnov distance between the cumulative distribution functions of the involved pixels, computed on the available frames, is evaluated. The method has been tested on simulated data and compared to other state of art despeckling filters belonging to different families, showing interesting performances in combining good details preservation with effective noise reduction.