Computer vision: image shape geometry and classification
Loading...
Date
2018-09
Authors
Pham, Dat Hoang
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This research introduces the study of shape analysis with shape descriptors based on the geometry of image triangulation. The Delaunay approach is used to superimposes on an image with a mesh filled with various sized triangles that give rise to various simplexes. In its simplest form, a simplex is a collection of path-connected vertices. The motivation for this approach in image analysis is that Delaunay triangulation covers image shapes having unknown geometries with simplexes that have known geometries that we can measure and compare. This approach provides the foundation for the study of image shape geometry and the extraction of features for many applications in computer vision such as image processing, image segmentation, object recognition and classification. Shape descriptors are constructed with features extracted from the geometry of the simplexes covering planar images. In this research, shape descriptors provide a framework for tracking the persistence tolerance of a shape over sequences of image captured my camera. An application that can be benefit from this approach is video content retrieval and classification
Description
Keywords
Computer vision, Content-based image retrieval(CBIR), Description, Feature vector, Shape descriptor, Persistence, Classification, Delaunay triangulation, Similarity measure, Persistence measure, Shape analysis, Video content retrieval.