摘要
提出了一种在含噪图像中检测边缘的快速算法。通过分析噪声的导数与边缘的导数的分布特点,推导出了三类模板算子:差分算子、邻域积分算子和非极值抑制算子。差分算子用于突出边缘,邻域积分算子用于抑制噪声,非极值抑制算子用于提取准确的边缘点。与Canny算子的对比实验表明,在弱噪声下,检测结果两者相当;在强噪声下,结果优于Canny算子,并且计算量仅为Canny算子的40%。
A novel algorithm for detecting edge from image with noise is proposed. Three template operators, difference operator, neighborhood integral operator and non-extreme value suppressor, are derived by analyzing the feature of distribution for noise differentials and edge differentials. Difference operator can emerge edge zones. Neighborhood integral operator can suppress noise. Non-maximum value suppressor can extract single edge from the edge zones. Experiments by comparing with Canny operator show that the result of edge detection using the algorithm is as good as using Canny operator for image with weak noise and is better than using Canny operator for image with strong noise. The cost of computation of the algorithm is just 40% of that of Canny operator.
出处
《科学技术与工程》
2008年第22期6028-6033,共6页
Science Technology and Engineering
