基于多尺度数学形态学梯度的图像检测

摘要

传统边缘检测算子如Canny、Sobel等是通过计算图像中局部小区域的差分来工作的,对噪声比较敏感,在检测边缘的同时常会加强噪声,形态边缘检测器在检测边缘的时候不会加强或放大噪声,单尺度形态学梯度算子结构元素过大或过小都不能检测到满意的结果.将多尺度形态学边缘检测和形态学滤波结合起来,提出一种边缘检测方法.首先利用形态学对多尺度的形态学梯度算子进行边缘检测,再进行填充内部缝隙,接着进行形态学边界平滑,最后通过分割掩模到原图像的结果.仿真结果表明,所提算法能够非常清晰地检测到目标图像的边缘,提取的图像边缘定位准确且平滑,同时具有较强的抗噪能力.%The traditional edge detection operators such as Canny, Sobel ere work by calculating small areas of the local image difference, more sensitive to noise, and enhance the noise while detecting image edge, form in the edge of the edge detector test time won＇t enhance or enlarge noise, single scaling morphological gradient operator structure elements have great or small all can＇t detect the satisfied result. In view of the shortage, multi -scale morphology edge detection and morphological filtering combined, put forward a kind of edge detection method. First, use the multiscale morphology of the morpho- logical gradient operator to detect the edges, and then fills in internal tiny cracks, then the boundary of image becomes continuity and smooth using the binary morphology, finally the result image is obtained through the segmentation masks. The simulation results show that the proposed algorithm are very clear to the target image edge detection, and the image edge extraction accurate orientation and smooth, and has strong antinoise ability.