关键词: 边缘提取, Gabor滤波器, 角点检测, 自相关矩阵

Abstract: To improve the accuracy of the corner detection, a new corner detection algorithm was proposed, and it used the Gabor directional derivatives of each pixel to construct the correlation matrix on edge contour.to detect corner. The algorithm firstly extracted the edge map of an image using the Canny edge detector; secondly, the image was smoothed by the Gabor filters and the correlation matrixes were constructed using Gabor directional derivatives of each edge pixel and its surrounding pixels. If the sum of the normalized eigenvalues was not only above the previously specified threshold but also the local maximum, it would be labeled as a corner. Compared with the traditional contour-based corner detection algorithm, it used the related information of the Gabor directional derivatives of the edge pixel and its surrounding pixels to extract corners, hence achieving better robustness to noise. The experimental results indicate that: The proposed algorithm detects more matched corners and fewer false corners in the noise free and noisy cases and achieves obvious improvement in performance.

Key words: edge extraction, Gabor filter, corner detection, autocorrelation matrix