Subpixel Corner Detection Using Spatial Moment

摘要

A novel subpixel corner detection method based on spatial moment is developed in this paper. Firstly, the spatial-moment-generating function, gradient magnitude and variation of gradient-direction corner are used as the decision rule of corner detection by analyzing the mathematical formula of corner-model spatial moment. Then Non-max suppression technique is utilized to detect the vertex of feature corner. Finally, in order to improve its localization performance, subpixel corner detection is implemented by the bilinear interpolation and Newton iteration method. Experiments illustrate that the spatial moment corner detector has better robustness and localization performance than Kitchen detector and Harris detector.