Aimed at presenting an approach to a range of problems in computer vision and machine learning that is data-driven, local and requires a minimal number of assumptions, this book shows how several problems can be posed as the organization of the inputs into salient perceptual structures, which are inferred via tensor voting.