摘要
从多模态图像配准的相似性测度中 ,选取具有代表性的统计型测度 ,从计算时间、对噪声的容忍性和图像窗口大小的影响等方面通过实验对它们的性能进行了分析比较 .考察的测度包括Shannon信息论测度、R啨ｎｙｉ熵测度、Tsallis熵测度、PIU测度及PIU改进形式 .文中结合多模态图像配准的实际情况 ,对PIU测度的数学表达式作了改进 .实验结果表明 ,改进的PIU测度性能有了提高 ,不同测度具有不同的有效性和适用范围 .
Typical similarity measures based on statistics are selected from those for multi-modal image registration. Then, the performance comparison and the analysis of these measures are made on the basis of computing time, tolerance for noise, and effect of image window size. These measures include Shannon information measures, Rényi entropy measure, Tsallis entropy measure, PIU(Partitioned Intensity Uniformity) measure and two improved formulae from PIU. The results of tests show that the improved PIU measures have the advantage over that of original PIUs.It is validated that these similarity measures have different performance and applicable cases.
出处
《计算机学报》
EI
CSCD
北大核心
2004年第9期1278-1283,共6页
Chinese Journal of Computers
基金
国家自然科学基金 (60 0 72 0 2 0 )资助
关键词
多模态图像
图像配准
相似性测度
信息理论
multi-modal image
image registration
similarity measure
information theory
