Echocardiographic image denoising using extreme total variation bilateral filter

Abstract The transthoracic echocardiographic (TTE) images used to assess cardiac health are inherent with speckle noise, making it very difficult for accurate abnormality diagnosis. To address this issue, a novel speckle reduction known as extreme total variation bilateral (ETVB) filter is proposed in this paper. The regularizer term of total variation (TV) method is replaced with the bilateral (BL) term in the proposed ETVB filter along with the prior term. The true information is incorporated in the algorithm using Bayesian inference and probability density function. Applications of gradient projection based restoration methods are also analyzed for speckle noise reduction. Denoising characteristics are evaluated in terms of 15 image quality metrics along with visual quality. The performance of proposed ETVB filter is compared with 30 existing despeckling techniques. Exhaustive result analysis reveals that the proposed ETVB filter is superior in terms of edge and texture preservation. The focal points of result analysis are edge, structure and texture preservation along with visual outlook. Edge and structure preservation are measured using beta metric, figure of merit and structure similarity index. The values of β, FoM and SSIM are markedly enhanced using proposed filtering scheme in comparison to other total variation based methods.