Abstract

Analyses that rely on the statistics of image neighborhoods have been shown by Buades-Coll-Morel, Awate-Whitaker, and many others to be very powerful for image denoising, reconstruction, and segmentation. The algorithms derive from several different formulations, all of which are based on nonparametric estimates of joint probabilities. In this talk we will discuss the various ways in which neighborhood statistics can lead to useful algorithms and describe several variational formulation of these statistical models that help explain the underlying potential of these methods.