Abstract: In order to increase precision and efficiency of Harris corner detection algorithm, this paper takes advantage of the modern Graphics Processing Unit(GPU) to improve Harris corner detection algorithm by parallelizing. A fast sub-pixel Harris corner detection algorithm is presented based on GPU. This fast algorithm combines the parallel characteristics of GPU with Harris corner algorithm. Experimental result shows that GPU can implement the real-time sub-pixel Harris corner detection for 24 bit 720×720 video images.

Key words: sub-pixel, corner detection, Graphics Processing Unit(GPU)

摘要： 针对Harris角点检测精度和检测速度问题，利用现代图形处理器(GPU)对角点检测算法进行改进，提出一种基于GPU的快速亚像素Harris角点检测算法，该算法利用了GPU的并行处理能力和亚像素Harris角点检测算法的并行性特点。实验结果表明，对于分辨率为720×720的24 bit视频图像，该算法能够实现实时的亚像素级Harris角点检测。

关键词: 亚像素, 角点检测, 图形处理器

CLC Number: 

• TP391