关键词: 边缘检测，轮廓检测，频域阂值分割，空域阂值分割

Abstract: Standard edge detectors compute the binary edge maps by thresholding the gradient image only in the spatial domain. When they are applied to the natural images, their results contain a lot of spurious edges because the gradient magnitudes of spurious edges are often stronger than that of object contours. To improve their performance in detection of object contours in natural scenes, we proposed a novel contour detection method by thresholding the gradient image in spatial-frequency domain in this paper. Taking Canny edge detector as an example, we used natural images with associaled subjectively defined desired output contour maps to evaluate the performance of the proposed method. Experimental results show that the proposed method can effectively improve the contour detection performance of the Canny edge delector.

Key words: Edge detection,Contour extraction,Thresholding in spatial-domain,Thresholding in frequency-domain