A Fast Computing Model for Despeckling Ultrasound Images

Abstract

Ultrasound imaging is a highly preferred diagnostic method due to its non- invasive nature; however, the presence of speckle noise degrades the quality of the images captured by this modality. Among the plentiful researches that happened in the field of despeckling, the non-local total variation methods have demonstrated promising results by maintaining relevant details and edges present in images. Nevertheless, the model is computationally expensive as it has to deal with large size matrices in computing the results, which in turn restricts its applicability in real-time environments. This study contributes a fast and numerically stable Non-local Total Variation Model for despeckling Ultrasound images. Multi-core GPU processors are employed for computing the parallelized algorithms developed using a fast converging Split-Bregman iterative scheme. A comprehensive evaluation is performed on the basis of execution time to demonstrate the efficiency of the model. © 2021, Springer Nature Singapore Pte Ltd.