面向光学薄膜瑕疵检测的二值图像快速Blob分析算法

针对光学薄膜瑕疵检测应用中的高实时性图像处理要求,提出一种基于游程的二值图像快速Blob分析算法。采用步进式动态扫描方式,每个游程仅需扫描一次,且不必与相邻行的所有游程进行比较,算法的搜索空间得到压缩;游程连通性比较的分支少,简化了判断过程,提高了操作效率;所设计的游程及目标对象的数据结构允许由任一游程节点快速访问其所属链表的首部和尾部,不仅为后续的数据访问提供了便利,且提高了标记冲突时链表合并的操作速度,避免了冲突等价表的介入。实验结果表明该算法具有鲁棒、高效的特性,并已在光学薄膜瑕疵检测系统中得到了应用。

Efficient Blob analysis of binary image for defects inspection in optical films

One of the key issues for a vision system used for the inspection of defects in optical films is the real-time performance request for image processing. A new Blob analysis algorithm based on Run-Length Encoding (RLE) for binary image was proposed, without any need to build an equivalence table or to unite equivalent labels. Since a dynamic step-by-step search method was adopted, each RLE needed to be scanned only once and did not need to be compared with all the RLEs at the upper line, which benefited a compressed search space. Quite limited branches were involved when analyzing RLE's connectivity, which simplified the process and improved the entire efficiency. With the specific data structure design, access of the header and the tail of a RLE list could be done directly, which made it efficient when handling labeling-conflict. The experimental results show that the proposed algorithm is robust and efficient. Moreover, a comparison of the computational results with the existing methods indicates that the proposed approach is a viable alternative for effectively labeling connected component. The algorithm has been successfully applied to a vision system for the inspection of defects in optical films.