Two-Step Noise Variation Estimation Based on Image Segmentation

Noise variation is an important parameter for many image denoising algorithms. This paper proposed a new two-step noise variation estimation algorithm based on image segmentation. In the first step, noisy image was smoothed and segmented by statistical region merge (SRM) algorithm. Then variation of every region was computed, and some regions were selected based on statistical rule to estimate the noise variation. In the second step, filter, segmentation and estimation parameters were revised according to estimated noise variation, and a new cycle of image filter, segmentation and estimation were performed to obtain more accurate estimation. Experimental results on large numbers of images and various noise show that, the proposed algorithm can estimate noise variation quickly and accurately. Its overall performance overcomes algorithms based on principal component analysis and singular value decomposition.