噪声对三维图像归一化互信息配准的影响  

Influence of Noise to the Registration of Three-dimensional Image Based on Normalized Mutual Information

作　　者：余慧婷[1] 张杰[1] 潘萌[2] 

机构地区：[1]暨南大学物理系,广东广州510632 [2]暨南大学光电工程系,广东广州510632

出　　处：《中国医学影像学杂志》2011年第11期844-849,共6页Chinese Journal of Medical Imaging

摘　　要：在三维图像配准方法中,归一化互信息法精度高、鲁棒性强、使用范围广。本文以归一化互信息为相似度对三维图像进行了配准,并采用了蒙特卡罗(MC)的方法研究噪声对三维图像配准的影响。实验发现,噪声的存在会造成归一化互信息的统计误差,使归一化互信息变小。为了消除噪声的影响,根据图像的灰度直方图,本文提出一种将灰度直方图进行区域分割后再配准的方法,减小了噪声造成的灰度值涨落和统计误差。研究表明,该方法可抑制噪声的影响,改善局部极值,减小误匹配,从而提高了归一化互信息,进一步完善了基于归一化互信息的图像配准。Image registration based on normalized mutual information is widely used because of its high accuracy and robustness. In this study, Monte Carlo method （MC） was used to study the influence of noise to the registration of three-dimensional image based on normalized mutual information as the similarity measure. The research results inferred that noise may lead to statistical error of normalized mutual information and decrease normalized mutual information. A novel method that the gray histogram of image was region-dependently segmented before registration was adopted to reduce the influence of noise to the registration. It decreased the fluctuation of grey and statistical error caused by noise. The experimental results showed that the region-dependent segmentation of gray histogram of image restrained the influence of noise, reduced local extrema and misregistration, increased normalized mutual information and improved the registration based on normalized mutual information.

关 键 词：归一化互信息 图像配准 区域分割法 

分 类 号：R445.9[医药卫生—影像医学与核医学]