摘要
文中提出了一种新的基于局部方差与残差复杂性的相似性测度.传统的基于灰度的相似性测度易受噪声、灰度偏移场和造影剂的影响造成误配.残差复杂性在一定程度上可以克服这一难点,但该测度对初始参数非常敏感,参数设置不正确往往达不到好的配准效果.文中利用图像的局部方差信息构造权重函数,在图像残差比较大的地方给予小的权重约束,在残差比较小的地方给予大的约束,计算约束后残差图像的残差复杂性作为新的相似性测度.新测度更平滑、鲁棒性更好,不容易陷入局部极值.对模拟数据和真实数据的实验表明新测度对噪声、灰度偏移场、造影剂和初始参数具有更高的鲁棒性,更加适合于医学图像配准.
We propose a new similarity measure using local variance and residual complexity.Traditional intensity-based similarity measures are easily disturbed by noise,intensity bias field and contrast agent.Even though residual complexity can tackle this problem in some way,it may not have robust performance due to initial parameter.To address the poor robust problem,our new measure employs local variance of reference image to construct weighting function.This function could automatically constraint the residual image.It gives small weighting value to large residual value,and vice versa.Then,we calculate residual complexity of constrained residual image.We validate our algorithms using both simulated data and clinical data.The experiment results indicate that new measure is more robust to initial parameters,noise,intensity bias field and contrast agent.
出处
《计算机学报》
EI
CSCD
北大核心
2015年第12期2400-2411,共12页
Chinese Journal of Computers
基金
国家自然科学基金(31000450)
广东省自然科学基金(2014A030313316)
广州市珠江科技新星专项基金项目(2012J2200041)资助
关键词
局部方差
指数函数
残差复杂性
鲁棒估计
图像配准
local variance
exponential function
residual complexity
robust estimation
image registration
