Neurocomputing and probabilistic propagation in computer vision

Abstract

One of the earliest (also well-studied) research areas in artificial intelligence is the study of visual perception, and the study of neurons of the brain using connectivist models or neurocomputing. Where cognitive and mathematical psychology, and neuroscience studied how the brain and perception works in their own paradigms, artificial intelligence provided tools from theoretical and applied computer science to study the aforementioned areas using digital computers. This study focuses on examining two sides of neurocomputing, namely probabilistic graphical models and artificial neural networks in solving early perception, or early vision and inference tasks. More specifically, the study examines probabilistic propagation such as denoising tasks under similarity measures and parallelization schemes. And finally, combining probabilistic graphical models and artificial neural networks into a pipeline model for solving inference tasks from a set of imaging measurements. Keywords: Algorithms, Artificial intelligence, Inverse problems, scientific computing.