Abstract: Edge detection is an essential pre-processing work in image segmentation and pattern recognition.The traditional methods such as Sobel，Prewitt，Canny and so on are useful and efficient but are sensitive to noise.Mathematical morphological edge detector is a heated researched field now.As it is known that general morphological methods adopt a single structure element.So they are difficult to detect complex edge feature.A novel multi-structure elements multi-scale morphological edge detection algorithm is proposed.Firstly the noise can be filtered by the alternative-order morphological open-close filter with a multi-scale element.Then different edge detection results by using multi-structure element morphological algorithm are extracted，and the final edge result is got by using synthetic weighted method.The experimental results showed that the proposed algorithm can filter noise successfully and is efficient for complex edge detection.

Key words: edge detection, mathematical morphology, multi-structure, multi-scale, binarization

摘要： 边缘检测是图像处理与模式识别的一个重要图像预处理过程。传统的边缘提取方法如Sobel、Prewitt和Canny等非常有效但对噪声非常敏感。形态学边缘检测目前已成一个研究热点，但大多算法采用单一结构元素，很难对复杂边界进行有效的处理。因此提出一种基于多结构元素多尺度的数学形态学边缘检测算法，先用多尺度结构元素交替顺序形态开-闭平滑图像以去噪，再用多结构元素对不同方向的边缘进行提取，最后把各方向边缘融合得到图像边缘。实验结果表明，提出的算法不仅有很强的抗噪性，而且很有效的提取图像的边缘。

关键词: 边缘检测, 数学形态学, 多结构元素, 多尺度, 二值化