Abstract: To take full advantage of multi-core processors hardware resource and computing capacity for improving the real-time of the image matching, by analyzing the theory of adaptive threshold sequential similarity detection algorithm, the related parallel algorithm, based on the multi-core and parallel programming model of task decomposition, is designed and realized by use of OpenMP programming model on multi-core computers, and then optimized. Experimental results show that the optimized parallel programs also increase the match speed, improve the multi-core utilization and achieve the expected effects well, in addition to keep the image match accuracy.

Key words: multi-core parallel, OpenMP, adaptive Sequential Similarity Detection Algorithms（SSDA）, image matching

摘要： 为了充分利用多核处理器的硬件资源和计算能力来提高图像匹配应用的实时性，通过对自适应阈值SSDA图像匹配算法原理的分析，基于任务分解的多核并行编程模式思想，设计了一种自适应阈值SSDA图像匹配并行算法，并在多核计算机上采用OpenMP模型编程实现该并行算法，同时还进行了相关的代码优化。实验结果表明，优化后的并行算法在保持匹配算法精度的同时大大提高了匹配速度和多核利用率，取得了良好的效果。

关键词: 多核并行, OpenMP, 自适应序列相似性检测算法（SSDA）, 图像匹配