Font Size: a A A

Multi-scale Representation Of The Two-dimensional Image Contours Method

Posted on:2011-10-05Degree:MasterType:Thesis
Country:ChinaCandidate:F HuFull Text:PDF
GTID:2208360302999667Subject:Applied Mathematics
Abstract/Summary:PDF Full Text Request
With the continuous development and advancement of science and digital image technology, people pay increasingly attention to the applications of the digital image processing and less attention to digital image representation. Digital image representation, however, directly impacts on the progress and development of other sectors of digital imaging technology. No matter how advanced digital image procession will be, the representation of image which has basic effects still has a relatively broad space for development.This paper will focus on the representation using geometric characteristics of the digital images. Basing on the analysis of existing, relatively mature technology Curvature Scale Space image representation (CSS), we study deeply the Triangle-area Representation and give our own views in the following aspects:(1) Experimentally analyzing the uniqueness and affine invariance of the Triangle-area Representation of image.The Triangle-area Representation, one of many geometry-based image retrieval methods, is the latest one with more matching accuracy and less computational complexity. In this paper we analyze the uniqueness and affine invariance of the Triangle-area Representation (TAR) for 2D closed-contour shapes. The experimental results show that the TAR can't satisfy uniqueness under affine transformations, which provides a theoretical basis for the further improvement of the Triangle-area Representation.(2) Initially proposing a novel technique for corner detection using Triangle-area RepresentationThe TAR method and CSS technique share the same basis, i.e. both of them represent image boundaries by the curvature under different scales. Nevertheless the CSS has been gradually perfected and extended in many fields, particularly in corner detection. Inspired by the CSS, we apply the TAR to the corner detection and then put forward a new algorithm. This corner detector is given through the Triangle-area Representation method, whose main idea rests on the maximum property of the local area at the corner point. Therefore, the method can be realized easily and is of high accuracy under some appropriate scale (triangle-length scale). Experimental results show that the new algorithm is feasible and will accelerate the improvement and development of the TAR. To some degree, the TAR corner detection operator provides a new relatively simple standard for prejudgment and posttest of corners.
Keywords/Search Tags:Digital Image Processing, The Curvature Scale-space Method (CSS), Triangle-area representation (TAR), Corner Detection
PDF Full Text Request
Related items