Denoising of medical ultrasound images based on non-local similarity: A low-rank approach

Abstract

Ultrasound imaging is one of the safest technique of clinical diagnosis. The presence of speckle noise has made denoising of ultrasound images indispensable for the proper diagnosis of diseases. Non-Local similarity and low-rank approaches is an upcoming area of research in the field of image diagnosis. However, their advantages have not been exploited in the denoising of the ultrasound image. In this work, a technique to denoise the ultrasound images by exploiting the low-rank property of patches grouped together is proposed. The patches are grouped utilising the idea of simultaneous sparse coding technique. This is followed by the singular value decomposition of similarly grouped patches which can be utilised for obtaining local and non-local information between the patches. The singular values are thresholded to obtain the denoised image. The proposed algorithm is compared with the state-of-art techniques existing in the literature and is seen to give the best results regarding both quantitative and qualitative measurements.