An Improved Harris Corner Detection Algorithus

Corner point is an important characteristic of the image. The corner point contains alot of information of the image, at the same time, the number of corners is far less than thenumber of pixels. We can master the shape of the objects in the image by calculating andanalyzing the corner points. Because the corner point has rotation invariance, it focused alot of important information on the image. Under the condition of no loss of imageinformation, the corner point minimizes the amount of data. Therefore, how to extract rightcorner form an image is the top priority of the work.Firstly, this article detailed introduces the corner detection, then analyses severalclassic algorithms. The author found in traditional Harris algorithm, the normalized imagecontrast is poorer; Harris operator’s ability to extract obviously corner point is weak; Susanoperator is more complex in threshold value. Against the above, this paper has adoptedthe following two solutions:1. Modify the constant k, through the theoretical demonstration and practical experiments,to enhance the visibility of the normalized image.2. Combine the Susan operator and improved Harris operator, using the union operator totest the corner of the image point, which can better play to their strengths.Through the changes of K, we enhance the image contrast, which can extract moreconvenient corner; narrow their choices of improved Harris algorithm’s threshold, whichcan make the algorithm find exact corner points in a smaller threshold range; using Susanalgorithm and the improved Harris algorithm to test the corner points of the image, canaccurate to the image corner detection in the noise interfered.The article finally carries on the summary to the full text, and put forward the need toimprove the algorithm. According to the actual application, corner radius r needs to bechoosen, and shorten the detection time as far as possible.