中文摘要：

针对复杂背景下的显微图像中的颗粒物检测，提出了一种快速、准确的颗粒图像分割新方法；该方法利用Canny算子得到图像中的颗粒边缘，采用格雷厄姆法计算颗粒边缘的凸壳，得到局部闭合区间；通过对单个闭合计区间内像素值信息的统计，得到局部阈值，利用局部阈值进行阈值分割得到颗粒二值图像；实验结果表明该方法有效克服了背景中纹理的影响，准确地分割出了颗粒图像，基于该算法开发的接触式表面洁净度检测仪具有较高的检测精度，得到了较好的应用。

英文摘要：

Aiming at detecting particle with complex background in micrograph, we proposed a new binarization algorithm which is quick and accurate at image segmentation. This algorithm uses Graham algorithm to get the convex hull of the edge and construct a local closed in- terval. By analyzing the information of pixels, the local threshold value could be gotten. Then, the particle image could he segmented using the local threshold in the closed interval. Experiments reveal that the particle image can be segmented effectively and accurately even if it is full of textures. Using this algorithm, we developed the surface cleanliness detectors.