多模态医学图像外边界点云数据实时配准仿真

Multi-Modal Medical Image Boundary Point Cloud Data Real-Time Registration Simulation

为了提升医学图像识别质量,需要对医学图像外边界点云数据进行实时配准。针对当前多模态医学图像外边界点云数据实时配准中,存在着配准精度较低、完成时间过长、误差较大等问题。提出基于特征点对齐度的图像外边界点云数据实时配准方法。采用SIFT算法提取出医学图像待配准特征点,以小波边缘检测法获取待配准特征点中的医学图像外边界点云数据,构建附近区域外边界点云数据特征,求出具有显著外边界点特征的特征点,结合Shape-context算子和显著外边界特征点,构建特征描述向量,利用角度直方图提取以特征点为中心的图像外边界点特征子图,计算出所有外边界点特征子图的对齐度来确定候选配准点对,引用线性加权法消除错误配准,以此完成正确配准。实验结果表明,所提方法在进行医学图像外边界点云数据配准中,配准精度较高、所需完成时间较短误差较小。

In this article,a method of real-time registration for cloud data of external boundary point in medical image based on the alignment metric of feature points was presented.Firstly,SIFT algorithm was used to extract the feature points to be registered in medical image.Secondly,wavelet edge detection method was used to obtain the cloud data of external boundary points in the feature points to be registered.Thirdly,the cloud data features of external boundary points in nearby region were established to get the feature point with remarkable external boundary point feature.Combined with Shape-context operator and significant external boundary feature point,the feature description vector was constructed.In addition,the angle histogram was used to extract sub graph of external boundary point feature centered on the feature point.Meanwhile,the alignment degree of all the external boundary point feature sub graphs was calculated to determine the candidate matching point pairs.Finally,the linear weighting method was used to eliminate the wrong registration,so as to complete the correct registration.Simulation results show that the proposed method has higher registration accuracy,less completion time and less error during the registration of cloud data of external boundary point in medical image.