ITERATIVE METHOD FOR BLIND EVALUATION OF MIXED NOISE CHARACTERISTICS ON IMAGES
Victoriya Abramova, Sergey Abramov, Volodymyr V. Lukin · Information and Telecommunication Sciences · 2015
A new method for blind estimation of mixed noise parameters is proposed. The method is based on line fitting into a set of cluster centers obtained from scatter-plot of local variance and mean estimates. Improved estimation of cluster centers is performed on basis of fourth-order statistical moment analysis. The estimation results for the proposed method are compared to the results for other known methods using images from TID2008 database. It is shown that the proposed method provides estimation accuracy comparable to the estimation accuracy of the method based on maximum likelihood estimation of image and noise characteristics (which is considered the best among the existing methods). An advantage of our method is that it is considerably faster.