使用传统的归一化互相关模板匹配算法进行肺结节检测耗时较长，在数据量较多的情况下容易造成漏诊或误诊。提出一种改进算法，主要从优化搜索策略入手，采用粗-精匹配思想，先使用改进SAD算法进行粗匹配找出侯准匹配点，再采用归一化互相关算法在侯准匹配点邻域内进行精确匹配找出最佳匹配点。实验结果表明，与NCC算法和卷积算法相比较，该算法在保证匹配精度的前提下，较大幅度地提高了匹配速率。采用这种算法进行自动肺节点检测可减少检测时间，对辅助完成早期的疑似肺结节点的定位和跟踪诊断有重要意义。%In the process of lung nodules detection, because of the large data, the conventional Normalized Cross-Correlation (NCC)algorithm will cost overmuch time, which increases the probability of missed diagnosis. An improved template matching method is proposed to solve the big time-consuming. This new method directs at the search strategy optimizing and adopts the coarse-fine matching idea. An improved Sum of Absolute Differences(SAD)algorithm is used to find all possible matching points firstly, then the best match point in the neighborhood of all possible matching points is found by adopting NCC algorithm. Experimental results show that Compared with NCC algorithm, this new method can not only ensure the accuracy of the template matching, but also greatly reduce the matching time. By using this algorithm, the time of lung nodules detection can be reduced and which is important for the location and tracking of lung nodules in the early stage.
展开▼
