Image feature corner extraction method based on Harris operator

The invention relates to the field of image splicing and recognition. On the basis of the original corner feature extraction method based on a Harris operator, two improvements are proposed: (1) a newcorner is generated if the absolute value of the difference between the transverse and longitudinal coordinates corresponding to two adjacent corners is smaller than a set threshold, and false corners are removed and the matching speed of images is improved by reducing the number of feature corners; and (2) mean processing is carried out on the gradient of grayscale within a certain range of image pixels, in order to reduce the corner information contained in the image grayscale gradient information and improve the speed of corner recognition. 本发明涉及图像拼接、识别领域。在原有Harris算子的角点特征提取的方法基础上提出了两项改进：1、临近的两个角点如果其所对应的横坐标和纵坐标差值绝对值小于个所设阈值，并通过相应的生成个新的角点。通过这种减少特征角点的方法，去除伪角点，提高图像的匹配速度。2、通过对图像像素定领域范围内的灰度的梯度做均值处理，减少图像灰度梯度信息中包含的角点信息，提高角点的识别速度。