基于连通算子的图像颗粒分割算法研究

An Algorithm to Separate Overlapping Particles Based on Connected Operator

为实现图像中粘连物体的有效分离,可以利用二值化分离背景,距离变换得到距离灰度图,形态重构提取每个物体的标识,分水岭算法确定边界。文中给出了实现的步骤及改进措施:提出基于双队列的方法来简化计算量,二次距离变换来提高分水线定位精确性,并在实际应用中实现。

In order to efficiently separate the overlapped particles in the image, some techniques have been developed such as binary conversion, distance function, morphology grayscale reconstruction and watershed transformation. Besides the introduction of theory and effect of every algorithm in segmentation, some improvements are provided to reduce the computing complexity and increase the boundary accuracy, including making use of two queues. Some examples are given to validate the applicability and good performance of the proposed algorithm.