Recovering surface brightness and depth from focus using the Expectation-Maximization algorithm

The problem of recovering surface brightness and depth from focus from a set of defocused images is discussed as an inverse problem and a new method is introduced. Focusing has long been noticed as an important cue for human and computer vision. It is becoming more and more important because it is a valuable alternative or complement to other classical cues like stereo or movement, as well as a critical information in the accomodation process. Our method is based on a version of the Bayesian Expectation-Maximization (BEM) algorithm, including the concept of deformable priors. It is shown that this formulation allows the recovery of meaningful surface estimates of a three dimensional scene without any assumption about the image contents. 1 1 Introduction Computer Vision is already a well defined scientific discipline, with a clear objective: the construction of systems that perform certain (intelligent) tasks based on visual sensing and feedback. The first processing stage in computer..