空域中基于正则化的自适应超分辨率图像复原  

Adaptive super-resolution image restoration based on regularization in spatial domain

作　　者：王晶[1] 彭国华[1] 

机构地区：[1]西北工业大学理学院,陕西西安710072

出　　处：《计算机工程与设计》2008年第12期3194-3196,共3页Computer Engineering and Design

基　　金：国家自然科学基金项目(60672135)

摘　　要：对于模糊图像的复原问题,从正则化技术克服问题病态性的思想出发,研究了一种有效的超分辨率图像复原方法。在Nguyen等人的正则图像复原框架的基础上,根据Roberts交叉梯度算子构造正则项,从自适应的角度生成正则化参数,并用共轭梯度法求解该模型的目标泛函极小值。计算机仿真结果表明,该方法可较好的再现图像的重要信息,复原图像的相对误差降低,同时,峰值信噪比和主观视觉效果方面都有明显的提高。Aiming at the restoration of blurred image, an effective approach to super-resolution image restoration is proposed based on regularization technique for dealing with ill-posed problem. Under the frame of the regularized super-resolution restoration which is proposed by Nguyen. Regularization item is choosen which based on the Roberts cross gradient operator and determines the regularization parameter adaptively. The conjugate gradient method is used to solve the minimum of the constraint equation, Computer simulations show that this proposed approach can properly retrieve the main information of original image, and the RE （relative error）, the PSNR （peak signal to noise ratio） and subjective effect of the restored image are improved significantly

关 键 词：图像复原 超分辨率 共轭梯度法 正则化 自适应 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]