A Generic Transfer Function based Technique for Estimating Noise from Images

of noise from an image continues to be a challenging area of research in the field of image processing. However, noise estimation from images that inherently contains very fine details or which have textured regions is still a challenging task. This paper attempts to estimate noise content from images corrupted by Gaussian and Speckle noise. The noise estimation technique proposed here is based on deriving a generic transfer function. This transfer function attempts to map the median value of the local noise standard deviation that is calculated on overlapping sub-images of the noisy image to the overall noise deviation in the image. The results obtained show that the proposed algorithm performs well for different types of images and over a large range of noise deviation. Comparison with other known standard techniques in literature is also presented in the paper, which confirms that proposed method provides better noise estimation. The approach has been proven to work on images affected by speckle noise as well. Results for estimation of speckle noise are also presented in this paper.