摘要
A mutual information-based non-rigid medical image registration algorithm is presented. An approximate function of Hanning windowed sinc is used as kernel function of partial volume (PV) interpolation to estimate the joint histogram, which is the key to calculating the mutual information. And a new method is proposed to compute the gradient of mutual information with respect to the model parameters. The transformation of object is modeled by a free-form deformation (FFD) based on B-splines. The experiments on 3D synthetic and real image data show that the algorithm can converge at the global optimum and restrain the emergency of local extreme.
A mutual information-based non-rigid medical image registration algorithm is presented. An approximate function of Hanning windowed sinc is used as kernel function of partial volume （PV） interpolation to estimate the joint histogram, which is the key to calculating the mutual information , And a new method is proposed to compute the gradient of mutual information with respect to the model parameters. The transformation of object is modeled by a free-form deformation （FFD） based on B-splines. The experiments on 3D synthetic and real image data show that the algorithm can converge at the global optimum and restrain the emergency of local extreme.
基金
Supported bythe National Basic Research Programof China ("973"Program) (No2003CB716103)
Key Project of Shanghai Scienceand Technology Committee(No05DZ19509)
关键词
非强制性注册
直方图
计算机绘图
信息共享
non-rigid registration
free-form deformation （FFD）
Hanning windowed sinc
partial volume （PV） interpolation
