Optimum parameter estimation for non-local means image de-noising using corner information

Non-local means (a.k.a. NL-means) method for image de-noising averages the similar parts of an image to reduce random noise. The de-noising performance of the algorithm, however, highly depends on the values of its parameters. In this paper, we introduce a method for finding the optimum parameters, present a linear estimation for the h parameter, and demonstrate that the most important parameter in this method is almost independent of the image and depends only on the noise. We also show that the de-noising performance can be increased by using corner information of noisy image. Our modifications result in better de-noising performance at less computational cost.