移动队列规则耦合角度约束的医学图像匹配

Medical Image Matching Based on Moving Queue Rule and Coupling Angle Constraint

当前医学图像的特征匹配主要依靠像素灰度来完成,但是像素灰度对空间信息不敏感,当匹配图像之间存在灰度信息不均衡以及噪声干扰时,将导致误匹配率较高,对此,本文提出了一种基于移动队列规则耦合角度约束的医学图像匹配算法.首先,利用高斯金字塔模型对源图像进行滤波预处理,以减少源图像中存在的噪声等干扰;再利用Harris算子对预处理后的源图像进行特征检测,获取图像的特征点;然后,利用SURF(Speed Up Robust Feature)特征描述子,获取特征点对应的特征描述子.并通过尺度空间理论获取特征点集,通过将特征点集进行排序来形成队列,从而设计移动队列规则,完成特征点的匹配;最后,通过求取匹配特征点间的夹角,形成角度约束模型,对匹配特征点进行提纯,剔除伪匹配特征点,使得匹配准确度得以提升.从仿真实验结果与分析可见,在对医学图像进行匹配时,本文所提出的方法具有匹配精度高、鲁棒性能好等特点.

Currently,feature matching is finished mainly relying on pixel gray level in medical image matching methods,which results in rather high mismatching rate when the gray information is not balanced between the matched images,and noise interference exists because pixel gray is insensitive to spatial information.Therefore,a medical image matching algorithm based on moving queue rule coupling angle constraint is proposed in this paper.First,the source image is pre-processed with the Gauss pyramid model to filter the source image for reducing the noise in the source image.Next,feature points of the source image are obtained by using the Harris operator to detect the features of the pre-processed image.Then,the feature descriptors corresponding to the feature points are obtained based on the SURF（speed up robust feature）descriptors.The mobile queue rule is designed based on the scale space theory to obtain the feature point sets and sorting these sets to form queue for completing the matching of feature points.Finally,the angle constraint model is formed by finding the angle between the matching feature points to eliminate the pseudo-matching feature points for improving the matched feature points and matching accuracy.An analysis of the results of a simulation experiment shows that this method has high matching precision and good robustness when used in medical image matching.