Multiscale corner detection based on wavelet transform and scale-space theory

Abstract

Being the front-end processing of a large amount of computer vision and image processing systems, corner detection is useful and important in the sense that it has a great impact on the following processing and consequently the performance of the whole system. In this thesis, we focus our research on the contour based corner detection (CCD) methods and direct intensity computation based corner detection (DICD) methods. Five algorithms are proposed, which partially or wholly solve three problems, i.e., (1) incomplete information; (2) delocalization; and (3) multiple responses to higher order structures. For the CCD methods, the thesis provides an overview of the existing corner detection methods for contour images covering classi¯cation, comparison and performance evaluation.