Image Corner Detection Method

In computer vision, machine vision and image processing, feature extraction is one of the important direction, and angular point is an important local features of images, which determines the target image, therefore corner detection is of significance in the shape of the image matching, objectives and identify and describe the motion estimation, target tracking etc. The corner of image processing is infromative , which can provide enough constraint, reduction of the computation, thus greatly improvies speed. After analisising on several traditional corner detection algorithms, the author puts forward a new algorithm which is based on the corner detection algorithm.First, some traditional corner detection algorithms have been studied and analyzed by pionting out the advantages and disadvantages of them.Secondly, through the research of Harris' algorithm the autor comes to the conclusion: Harris' algorithm lacks accuracy. By using b-spline interpolation on local image of the image interpolation and the second corner detection the flaw of Harris' algorithm can be solved. Experiments prove that the improved algorithm can greatly improve the accuracy of the original algorithm, it can reach the accuracy of 1×10-1.Finally angular point extraction algorithm based on the template is proposed. By using the this algorithm the author comes to a new method to get the angular point automatically. According to the characteristics of the corner the author gets 3x3 templates, which contain all the possible modes of the image, and then code the templates. Before abstracting the corner a series of image processing have to be done, templates can be used to its efficiency. Experiment proves, compared with the traditional algorithm, this algorithm achieves better accuracy without human intervention.