A Non-local Maximum-Likelihood Denoising Algorithm
Matthew D. Sambora · 2007
Digital images obtained using CCD or CMOS sensors are subject to corruption by AWGN noise during sensor readout. Most of the techniques used for noise reduction in images have relied on methods that operate on global image attributes and noise assumptions. Recently, several approaches have been proposed that attempt to recover an un-corrupted image by examining attributes within a statistical neighborhood of an image. This paper will extend this body of work by describing a novel statistical neighborhood algorithm for denoising images. This algorithm exploits the naturally occurring redundancy in an image and employs an algorithm to selectively normalize and average redundant information in the corrupted image.