基于最小二乘准则的模糊估计和图像复原
Blur Identification and 3D Image Restoration Based on Least square Theory
作者：卿粼波(四川大学 电子信息学院,图像信息研究所，四川 成都610064)；何小海(四川大学 电子信息学院,图像信息研究所，四川 成都610064)；陶青川(四川大学 电子信息学院,图像信息研究所，四川 成都610064)
Author：(Image and Info. Inst.,School of Electronics and Info. Eng.,Sichuan Univ.,Chengdu 610064,China)；(Image and Info. Inst.,School of Electronics and Info. Eng.,Sichuan Univ.,Chengdu 610064,China)；(Image and Info. Inst.,School of Electronics and Info. Eng.,Sichuan Univ.,Chengdu 610064,China)
收稿日期：2007-06-21 年卷（期）页码：2008,40(2):129-133
期刊名称：工程科学与技术
Journal Name：Advanced Engineering Sciences
关键字：计算光学显微成像；最小二乘；模糊估计；图像复原
Key words：computational optical sectioning microscopy;least square;blur identification;image restoration
基金项目：国家自然科学基金资助项目（60372079）
中文摘要
在计算光学显微成像技术中，点扩展函数往往是未知的，且不易获取，从而给图像复原带来很大困难。基于最小二乘准则和最优化理论，提出了利用变尺度法的三维点扩展函数参数估计算法；针对传统EM算法存在复原效果细节丢失严重等问题，提出最小二乘共轭梯度三维图像复原算法。算法在点扩展函数参数估计和求解真实图像之间进行交替迭代，从而得到图像的最优估计。实验表明，新算法在较短时间内，能够较准确地估计出点扩展函数参数，并得到较好的复原结果。
英文摘要
The point spread function（PSF）in computational optical sectioning microscopy (COSM) technology was always unknown and hard to obtain. This brought a lot of difficulties to image restoration. Based on least square and optimal theory, a parameter estimation method using variable metric was proposed for 3D point spread function. In addition, to overcome the traditional EM algorithm’s limit such as serious detail loss, a conjugate gradient least square(CGLS) algorithm for the restoration of 3D images was developed. The optimal real image estimation was obtained through alterative iteration between the parameter estimation of point spread function and the real images estimation. Experimental results showed that the new algorithm could successfully identify blurs and restore images, in a short time.
【关闭】
