摘要
针对低分辨率图像盲复原中信息不足的问题,可以用正则方法来求解。假设点扩散函数结构已知而参数未知,模糊矩阵可表示为带参数的形式,在Nguyen等人的正则有参盲复原框架的基础上,进一步根据Roberts交叉梯度算子构造正则项,从自适应的角度构造正则化参数,并用迭代法求解该框架的目标泛函极小值。算法分析和实验结果表明,这种方法能取得令人满意的超分辨图像复原效果。
The problem of blind image restoration without sufficient information requires effective regularization to stable the solution supposed the point spread function is known while its parameters are unknown.Based on the frame of the parameters and regularized blind super-resolution restoration proposed by Nguyen,this paper introduces a regularization item with Roberts cross gradient operator and determines the regularization parameter adaptively.Besides,the conjugate gradient method is employed to search the minimum point.Experimental result confirms that the restored images posses low relative error and high quality.
出处
《计算机工程与应用》
CSCD
北大核心
2008年第6期42-44,109,共4页
Computer Engineering and Applications
基金
国家自然科学基金(the National Natural Science Foundation of Chinaunder Grant No.60672135)。
关键词
超分辨
盲复原
广义交叉准则
自适应正则
共轭梯度法
super-resolution
blind restoration
generalized cross-validation
regularization
conjugate gradient
