Speckle Filtering of Ultrasound B-Scan Images- A Comparative Study of Single Scale Spatial Adaptive Filters, Multiscale Filter and Diffusion Filters

TLDR: A cumulative speckle reduction (CSR) algorithm in the MATLAB environment, which performs all despeckle filtering functions as well as performance metrics calculation in a single trial is developed, which finds that SRAD and Wavelet despekling filters are exhibiting fairly well performance over the other standard spatial filters.

Abstract: This paper presents a comparative study of among one Multi Scale filter (wavelet), nine single scale spatial adaptive filters (viz. Anistropic Diffusion (PMAD)) that are widely used for speckle noise reduction in biomedical ultrasound B-Scan images. The main objective of this study is to identify the efficient and optimum speckle filter in terms of preserving the edge details of the images with effective denoising. The performance of the filters are best estimated by calculating twenty-one established performance metrics along with execution time in order to determine the effective and optimum-despeckling algorithm for real time implementation. To do this, we have developed a cumulative speckle reduction (CSR) algorithm in the MATLAB environment, which performs all despeckle filtering functions as well as performance metrics calculation in a single trial. In the case of diffusion filter implementation, provision is given to execute the diffusion filter for several trials to identify the best iteration in terms of denoising the speckle and preserving the diagnostic information found in the B-scan images. The algorithm has been experimented with more than 200 digital ultrasound B-scan images of kidney, abdomen, liver and choroids. Based on the visual inspection of the despeckled images and the calculation of the performance metrics, it is found that SRAD and Wavelet despeckling filters are exhibiting fairly well performance over the other standard spatial filters.