摘要：

针对Harris 算法在图像处理过程中特征点提取实时性和抗噪能力较差、计算量大的问题,提出一种结合像素点灰度差的Harris 角点检测算法。将被检测点与半径为3 的邻近圆周上16 个像素点作对比,以此计算非相似像素点的个数来确定候选角点,通过Harris 角点响应函数提取角点,并结合SUSAN 算法的思想去除伪角点。实验结果表明,改进算法提高了原算法的实时性,增加了角点提取的数量,并且能有效去除大多数伪角点,提高图像角点检测的速度和正确性。

关键词: Harris 算法, 实时性, 角点检测, 像素, SUSAN 算法, 伪角点

Abstract:

Against the issues of Harris algorithm for image processing for extracting feature points less real-time,the
calculation is larger,and sensitive to noises. This paper proposes a Harris corner detection algorithm combined with
pixels’ gray level difference. In this algorithm,the point is detected taking contradistinction from adjacent 16 pixels on
the circumference of a circle that radius equals to 3. The number of non-similar pixels is calculated and judged as a
candidate corner and then the Harris corner response function is calculated to extract corner,combing with the idea of
SUSAN algorithm to remove false corners. Experimental results show that the algorithm improves original algorithm’s
real-time while increasing the number of corner extraction and removing most of the false corners effectively,and helps to
improve the speed and accuracy of image corner extracting.

Key words: Harris algorithm, real-time, corner detection, pixel, SUSAN algorithm, false corner

中图分类号: 

• TP391