Blind noise level detection

In this paper we present a new fully automatic algorithm for blind noise level evaluation based on natural image statistics (NIS). Natural images are unprocessed reproductions of a natural scene observed by a human. During its evolution, the Human Visual System has been adjusted to the information encoded in natural images, making images interpreted best by a human when they fit NIS. The main requirement of such statistics is their striking regularity. Unfortunately, most computer images suffer from various artifacts, such as noise, that distort this regularity. Our contribution is applying the statistical behaviors for noise level evaluation. As most denoising algorithms require the user to specify the noise level automatization of the process makes it more usable and user independent. We compare the quality of our results to other algorithms.