Shape Representation Based On Graphs

With the rapid development of the multimedia and network technology, digital images have been extensively used in different disciplines. Content-based Image Retrieval becomes one of the most active research areas in multimedia retrieval field. The main idea of CBIR is to analyze the information of an image by the color, texture, shape, and other low-level image features, taking feature vectors as retrieval indexes and matching similarities based on multi-dimensional feature vector of images. Image feature representation is one of the key issues in CBIR. Shape depicts one of the most essential attributes of a physical object, and it is also one of the most difficult image features to describe. Moreover, obtaining three dimensional(3D) shape information from a sequence of images and reconstructing and displaying of the objects are as well the leading project of research. 3D shape representation and recovery can be used in a variety of applications such as object tracking, recognition, navigation, surveillance and visualization.Based on graph and graph spectral theory, this dissertation mainly investigates shape representation based on graph, including two aspects:(i) The research on the efficient methods of representing shape by Delaunay graph and graph spectrum, and (ii) the research on 3D representation and 2D shape display and rebuild based on octree(a special graph). The main contributions of the dissertation are outlined as follows.Firstly, the representation method of an object shape based on its relational graph and graph spectrum is investigated. The techniques of content-based retrieval are analyzed. Image retrieval based on shape is the main difficult part of CBIR. When we find the images or visual content of videos what we need, object...