融合确定性信息和随机信息的插值方法研究

Novel interpolation algorithm for integrating deterministic and stochastic information

为了提高医学图像配准过程中的测度曲线光滑性和运算速度,本文利用图像的灰度概率分布作为确定性信息,同时利用非整数网格位置处的灰度随机性信息,定义了融合确定性信息和随机性信息的置信区域(DSCR);结合最近邻域插值法,提出了基于DSCR的最近邻域插值法(DSCRNN)。使用DSCRNN插值方法得到测度在整数平移位置处的值是准确无误差的。通过医学图像之间的刚体配准实验,从函数曲线、运算时间、抗噪鲁棒性和收敛性能方面对比分析了8种插值方法,结果表明,相对其它插值方法,DSCRNN插值方法在不牺牲插值速度的前提条件下可以提高归一化互信息(NMI)测度的收敛性能和抗噪声能力。

In order to enhance the smoothness of the measurement curve and accelerate the registration speed in medical image registration,the confidence region integrating deterministic information and stochastic information（DSCR） is defined,where the deterministic information is the intensity probability distribution in images and the stochastic information is the stochastic intensity information at non-grid position.And then a new nearest neighbor interpolation method based on DSCR is proposed,which is abbreviated as DSCRNN.The values of normalized mutual information（NMI） are deterministic and accurate at any grid translation position when the DSCRNN interpolator is used.The measures′ curves,interpolation time,noise immunity and convergence are compared by applying 8 interpolation methods to the rigid registration of brain images.The results of tests show that the new DSCRNN interpolator outperforms the other interpolators in convergence performance and noise immunity without compromising interpolation speed.