基于广义变分正则化的闪光照相图像重建算法  

Generalized variation-based regularization algorithm for image reconstruction in high energy X-ray radiography

作　　者：钱伟新[1] 刘瑞根[1] 王婉丽[1] 祁双喜[1] 

机构地区：[1]中国工程物理研究院流体物理研究所,四川绵阳621900

出　　处：《强激光与粒子束》2009年第12期1903-1907,共5页High Power Laser and Particle Beams

基　　金：中国工程物理研究院双百人才基金项目(2008R0102)

摘　　要：针对闪光照相图像信噪比低的特点,提出了一种基于广义变分正则化的图像重建算法,该方法采用p-范数取代目前广泛采用的全变分范数作为正则项,构造了用于图像重建的展平泛函,将图像重建问题转化为目标泛函最优化问题,采用固定点迭代法求解图像重建的最优解。数值计算结果表明,该算法在重建过程中能够有效抑制图像噪声,并加大对图像边缘的保持能力,从而提高了图像重建质量,是一种有效且性能优良的闪光照相图像重建算法。According to the characteristics of flash radiographic image with low signal-to-noise ratio, a generalized variation （GV） regularization based image reconstruction algorithm is proposed. In the new algorithm, p-norm is used as regularized term instead of total variation（TV） norm in widely-used TV-based image denoising methods. Then a smoothing functional is construc- ted for image reconstruction. Thus, the problem of image reconstruction is transformed to a problem of functional minimization. A nonlinear partial differential equation（PDE） is deduced from the new image reconstruction model. To solve the nonlinear PDE, fixed point iteration（FPI） scheme is introduced to linearize the PDE, ensuring the stability and convergence of regularized solution. Numerical results show that, compared with TV regularized algorithm, the GV regularized algorithm can preserve image edge and suppress noise more efficiently while reconstructing image. Thus, the GV regularized algorithm is an efficient image reconstruction algorithm for flash radiography with better performance of noise smoothing and edge preserving.

关 键 词：闪光照相 广义变分 正则化 全变分 非线性偏微分方程 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]