一种改进边缘连接的Canny边缘检测算法

针对Canny边缘检测算法中需要手动设置固定阈值而出现边缘间断或伪边缘的问题,设计了一种基于边缘对比特征和边缘方向的边缘连接方法。该方法是基于视觉感知实验总结的一组边缘对比阈值数据比较强边缘点和待连接边缘点的对比特征判断其相似性,通过保持一定的边缘方向确保边缘连接方向的正确性,待连接点只有满足相似性且连接后新旧边缘方向一致,才能被连接为新边缘点。结果表明,改进边缘连接的Canny边缘检测算法具有很好的边缘连接能力,且能获取完整干净的边缘。

An improved edge linking algorithm for Canny edge detection

In light of edge discontinuity or pseudo-edge caused by manually set fixed thresholds with the traditional Canny edge detection algorithm, this paper proposes a new edge linking method based on edge contrast features and edge direction which judges the similarity between the stronger edge point and the to-be-connected edge point by comparing their contrast features based on a set of edge contrast thresholds summarized by the visual perceptual experiment. The method ensures the correctness of the edge linking direction by keeping a certain edge direction, and the to-be-connected edge point won't become a new valid edge point unless it meets the similarity and is in the same direction with the old edge point. The results show that the improved Canny algorithm for edge linking can lead to complete and clean edges with efficient edge connectivity.