A novel noise reduction method for optical coherence tomography images

In the field of image processing, compared with noisy image, high-quality or clean image usually has a positive impact on the final results, which lead to the emergence of image denoising. Recently, the patch-based methods have attracted considerable interest of researchers, which leverage the self-similarity of the noisy image or information of other relevant images. In this paper, we proposed a novel patch-based method to reduce the noise in optical coherence tomography (OCT) image, which consist of two steps. A combination of internal denoising and external denoising is adopted to select similar patches for each patch in the noisy image. Next, we use low rank technique to denoise the group of the noisy patch and its similar patches. After all patches in a noisy image are denoised, the corresponding clean image is constructed by aggregating all denoised pixels and the first step denoising is accomplished. A new noisy image is constructed and taken as the input of the second step denoising. Experimental results demonstrate that our method achieves better effect and performance, compared with the existing state-of-the-art methods.