An improved non-local means algorithm for CT image denoising

The non-local means (NLM) algorithm is a classical image denoising algorithm. However, the denoising effect of the NLM algorithm is easily affected by the noise level of neighboring pixels, which leads to poor denoising effect for high noise level image. In this paper, an improved NLM (I-NLM) denoising algorithm is proposed, which can extract the gradient information of the image more accurately by fusing the Laplacian of Gaussian (LOG) operator. At the same time, the algorithm combines the real domain information and the gradient information of the image to calculate the weight of the similarity between the image blocks. Experimental results show that compared with the traditional NLM algorithm, the proposed I-NLM algorithm can effectively preserve the edge of the image while suppressing the noise, and recover the CT images with high peak signal-to-noise ratio (PSNR) and structural similarity index measure (SSIM).