一种并行二值图像连通域标记算法

A Parallel Algorithm for Connected Component Labeling in Binary Images

连通域标记算法在CPU运行效率比较有限,这也是连通域标记算法处理图像的症结所在,针对这一问题,提出了一种并行二值图像连通域标记算法。从二值图像处理问题的可并行化出发,根据GPU并行计算的特点设计出了一种图像预处理、溯源(伪溯源)、伪溯源处理三个步骤的并行标记算法,采用硬件支持的多个线程并行执行处理的方式来提升算法的效率并实现了该算法,实验结果表明该算法相对CPU单线程处理算法效率具有明显提升,也更具有鲁棒性,尤其是图像像素不断增多后,效率更加明显。

The connected domain labeling algorithm has limited operating efficiency in the CPU.This is also the crux of the connected domain labeling algorithm for image processing.To solve this problem,a parallel binary image connected domain labeling algorithm is proposed.Starting from the parallelization of binary image processing problems,the article designs a parallel marking al⁃gorithm with three steps of image preprocessing,traceability(pseudo traceability),and pseudo traceability processing based on the characteristics of GPU parallel computing.The article uses hardware-supported multiple threads to execute processing in parallel to improve the efficiency of the algorithm and implement the algorithm.The experimental results show that the efficiency of the algo⁃rithm is significantly improved compared with the CPU single-threaded processing algorithm,and it is more robust,especially when the image pixels continue to increase,the efficiency is more obvious.