Time Scale Estimation Method Based on Anisotropic Diffusion

How to determine the stopping mechanism and parameters is the key in the anisotropic diffusion filtering of images. This paper introduces image structures similarity (SSIM) to time scale estimation based on anisotropic diffusion. It proposes a novel estimation method of the optimal diffusion time scale by SSIM. The method correlates well with the judgment of human observers and needs not to estimate noise variance compared with MMSE, SNR, minimal correlation coefficient and others. The efficacy of the scheme is demonstrated with experiments on some images.