首页 | 本学科首页   官方微博 | 高级检索  
检索        

基于Contourlet变换的CT和锥形束CT图像配准算法
引用本文:岳海振,李海云,刘迪.基于Contourlet变换的CT和锥形束CT图像配准算法[J].北京生物医学工程,2012,31(2):140-145.
作者姓名:岳海振  李海云  刘迪
作者单位:首都医科大学生物医学工程学院,北京,100069;首都医科大学生物医学工程学院,北京,100069;首都医科大学生物医学工程学院,北京,100069
基金项目:国家自然科学基金,北京市自然科学基金
摘    要:目的提出一种基于Contourlet变换,用于放射治疗定位的CT与锥形束CT(cone beam CT,CBCT)图像配准的方法。方法利用Contourlet变换多尺度多方向的分辨特性,将待配准图像进行Contourlet变换分解,分解后的高频方向子带合成梯度图像,采用归一化互信息作为相似性测度,把梯度图像与低频方向子带以加权函数结合,进行临床医学图像的刚性配准,有效弥补了互信息配准中缺少空间信息的不足。结果通过已知空间变换参数图像的配准结果验证了算法的准确性。配准后lO幅图像变换参数的误差极小,且均方根误差接近于0。结论该图像配准算法精确度高,并具有很好的鲁棒性,有助于提高图像引导放射治疗(image guid edradiation therapy,IGRT)中解剖组织结构和靶区的定位精度。

关 键 词:图像配准  多分辨率分解  Contourlet变换  互信息

Image registration algorithm in CT and cone beam CT based on Contourlet transform
YUE Haizhen,LI Haiyun,LIU Di.Image registration algorithm in CT and cone beam CT based on Contourlet transform[J].Beijing Biomedical Engineering,2012,31(2):140-145.
Authors:YUE Haizhen  LI Haiyun  LIU Di
Institution:School of Biomedical Engineering, Capital Medical University, Beijing 100069
Abstract:Objective A novel image registration algorithm in CT and cone beam CT (CBCT) is proposed for the localization in radiotherapy based on Contourlet transform. Methods The muhi-directional and multi- resolution Contourlet transform is applied to decompose the original image. The gradient image is constructed from the high frequency directional subbands,and normalized mutual informaion (NMI) is used as the similarity measures to calculate the mutual information of the gradient image and low frequency directional subband ,respectively. Then the synthesis similarity measure is the combination of mutual information for two images with a specific weight function The proposed approach can compensate for the lack of spatial information in the mutual information based on image registration. Results The algorithm accuracy is verified by comparing the registration results of ten medical images with specific spatial transformation parameters. The errors of the spatial transformation parameters after registration are small,and the mean squared error (MSE) is close to zero. Conclusions The experimental results are accurate and the algorithm is robust. This method improves the localization accuracy of anatomical structures and targets in the applications of image guided radiation therapy (IGRT).
Keywords:image registration  multiresolution decomposition  Contourlet transform  mutual information
本文献已被 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号