摘要
边缘检测的任务是寻求噪声平滑与边缘定位的最佳折中.本文基于一般梯度边界模型提出一种结合噪信比与定位精度的新的边缘检测准则,导出满足最佳性质的算子,它们可以利用递推计算高效实现。可以证明沈俊算子和Canny-Deriche算子分别是一、二阶最佳算子的定常系数线性特例,它们在新准则下是统一的,本文给出对测试和实际信号的实验结果.
Edge detection is to find the optimal trade-off between noise restraining and signal preserving. In this paper a new edge detection criterion involving noise-to-signal ratio and localization ability is proposed, based upon a general gradient edge model. The optimal operators can be implemented efficiently by recursive computations. It can also be proven that the first and second order optimal linear shift-invariant filters are Shen's and Canny-Deriche's operators respectively. Experiments involving both test and real images are available.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
1998年第1期54-61,共8页
Pattern Recognition and Artificial Intelligence
关键词
图像处理
边缘检测
自适应滤波
递推算子
Image Processing, Edge Detection, Adaptive Filtering, Recursive Operators
