基于二维图像形态处理法的接触物体分离算法

在许多图像分析处理中，区分和计算被摄对象的形状与尺寸大小是重要的一步、解决形状识别等问题对于图像中那些互不接触的被摄对象(本文中简称物体)而言是相对简单的．如果物体相互接触甚至部分重叠，问题就变得复杂起来．在这种情况下，2个或2个以上接触的物体相对机器视觉来说只是一个尺寸更大，形状更复杂的物体，由此得出的结果无疑是错误的．本文作者提出了一种利用图像形态处理法来分离相互接触或轻度重叠物体的算法．该算法首先对二值图像中的物体图形部分进行“三维地貌化”的立体化处理，再设计一个搜寻模板在所得的地貌表面上通过搜寻特征面来寻找分离点(鞍点)．分离点位置确定后，搜索最佳切割方向继而进行图形的切割分离．该算法在一含有相互接触物体堆的真实图像上进行验证，其结果显示了算法是有效而令人满意的．

An Approach of Separating Touched Objects Based on 2D Image Morphological Processing

In many image analysis and processing problems, discriminating the size and shape among the objects projected in an image is an important practice. It is relatively easy to distinguish these features among those objects already separated from each other. The problems will be undoubtedly more complex if the objects are touched or/and overlapped. The algorithm that can be used to separate the touches and overlaps existing among the objects in an image is presented. The approach is first to convert the gray-scale image to its corresponding binary image and then consequently to the 3D topographic one. A mask is engineered to search the topographic surface for the saddle point, from which the cutting orientation is determined followed by the separating operation. The algorithm is tested on a real image and the running result is quite satisfying and encouraging.