Image Corner Detection Method

Corners are important image features that can be used in object recognition,stereo matching,motion estimation and other field.Corner detection is the basic problem in the field of image processing,which is an important method of low level image processing. In this paper,several traditional corner detection algorithms are researched and the advantages and disadvantages of these algorithms are analyzed.According to these researches,a new algorithm of corner detection based on template is presented.Three parts are mainly developed in this paper:1.Research on traditional corner detection algorithm.According to studying several traditional corner detection algorithms and analyzing the advantages and disadvantages of these algorithms,we can get a conclusion that the accuracy of the Harris algorithm is not very high.Sometimes the accuracy can not meet the requirements.2.Study on improved Harris corner detector.Because the accuracy of the Harris algorithm is not high,B-spline is use to interpolate the local images and then corners of the local image are detected.After that sub-pixel coordinates of the corners are gotten.Experiments show that the new method is accurate in corner detection and average location errors of corner are about 1×10^-1.3.Study on corner detection algorithm based on template.Based on the template,a new algorithm of corner detection was presented.A 3×3 square template was designed, which includes all possible models of corners in image.These models were coded according to a specific rule and positions of these corners can be located by the template after denoising,swelling and thinning.Experimental results show that this algorithm can detect corner automatically without manual interposition,and average location errors of corner are at one pixel.