Image Denoising by using Non-Local Means and Total Variation

Abstract

Recently, medical modalities such as low dose CT, MRI and tomosynthesis have focused on generating noise-free images by using fewer measurements. However acquiring or using less data to reconstruct an image increases the noise level in the image. Thus, image denoising has been one of the most active research areas due to the noise existence in most medical imaging modalities. Due to its virtue of edge preserving, Total Variation (TV) has been actively used in medical imaging. Non-Local Means has recently been proposed as a filtering to suppress the Gaussian noise and preserve fine details in the image. In this study, the total variation (TV) minimization, is combined with Non-Local Means (NLM) filtering to increase the noise reduction. Visual and numerical results show that an important improvement in image denoising has been achieved in the sense of Structure Similarity (SSIM) and RMSE. The optimum NLM filtering parameters selection has also been studied to increase the performance the proposed method.