相干散射图像圆心估计中的交互式图像区块间互相关算法研究

Interactive cross-correlation algorithm among image blocks in circles center estimating for coherent scatter images

目的:解决X线相干散射图像中光谱圆环的真实圆心近似求解的问题。方法:通过交互的方式,在图像上选定一个估计的圆心。依据选定的圆心对图像进行对称分区,分区之间视为不同的图像块,计算图像块之间的相关系数。在估计的圆心周边选定一个邻域范围,重复上述的计算过程,分别计算出各个邻域点作为估计圆心时的相关系数。比较这些相关系数求出最大值,其对应的估计圆心即为近似的真实圆心。结果:使用理论模型模拟,真实圆心命中率约为99%。没有命中的最大偏差为3个像素,在光谱分析中是可以接受的。使用交互式图像区块互相关系数求解近似圆心的方法对实际的图像求解圆心,光谱的平滑度明显提升,波峰提升约为15%、波谷降低约为10%。结论:交互式图像区块互相关系数求解近似圆心的方法较好地解决了计算圆心和实际圆心的逼近问题,可以在实际工作中较准确地逼近实际圆心,提高光谱曲线的精度及分析的准确性。

Objective To solve the problem how to get the approximate solution of the real centroid of spectrum pattern circles in X-ray coherent scatter image.Methods A hypothesis centroid was selected in interactive mode,then this point was taken as the real centroid to segment the image into parts symmetrically.Each part was processed as an image.In case the hypothesis centroid was the real centroid,there was the maximum correlation coefficient between a pair of symmetrical image parts.Otherwise,the correlation coefficient would be reduced.With calculating the departure grade of a coefficient,the hypothesis centroid approaching the real centroid was corrected.Results The theoretical model showed that the real centroid had a 99%hit rate and the maximum departure distance was 3 pixels between the real centroid and the approximate centroid.This result was acceptable for the analysis of spectrum curve.With the approximate centroid calculated in the real image,the spectrum curve was promoted.The smoothness of the curve was improved obviously.The wave crest of the curve was raised about by 15%.The troughs of the curve descended about by 10%.Conclusion The departure of the approximate centroid from real centroid will bring a serious error into analysis result.It can influence the smoothness of spectrum curve,the zone of energy and the result of spectrum spreading.The method above solves the problem of the approximate centroid approaching the real centroid.The accurate centroid can improve the precision of the spectrum curve and the accuracy of the analysis results.