基于双参数正则化方法的图像复原  

Study of Blind Image Restoration Based on Two Factors Regularization Approach

作　　者：袁泉[1,2] 张骋[1,2] 陈建军[3] 严一清[3] 金聪[3] 李雪[3] 

机构地区：[1]重庆邮电大学,重庆400065 [2]重庆信科设计有限公司,重庆401121 [3]重庆大学通信工程学院,重庆400044

出　　处：《电视技术》2013年第15期1-3,共3页Video Engineering

基　　金：国家自然科学基金项目(61071190);中央高校基本科研业务费资助项目(CDJRC10160004)

摘　　要：针对高斯模糊图像的纹理和结构特性,提出一种基于总变分的双参数正则化约束的图像复原方法。首先采用基于各向异性的梯度处理方法,建立了具有点扩散函数和图像像素的灰度梯度。然后进行非线性和空间各向异性的规整化处理,使其在恢复目标图像和估计点扩展函数中能自适应地进行梯度运算。最后采用交替迭代的数值处理方法提高图像复原的速度。实验结果表明,基于梯度的双参数正则化图像复原能较好地提高图像复原的效果。According to the properties of image texture and image structure, the blind image restoration based on the two factors regularization approach is proposed. The total variation based on anisotropic and nonlinear regularizations are discussed in this paper. The anisotropic and adaptive regularizations are adopted according to the properties of turbulence point spread functions and image. The nonlinear and spatial anisotropic regularization functions are suggested to smooth adaptively the gradients in the process of estimating the point spread functions and recovering the object image. Finally, the cost functions are minimized by alternate minimization scheme. The result of the simulation demonstrated that the proposed algorithm improves the quality of the recovered image significantly.

关 键 词：正则化 图像复原 总变分 点扩散函数 自适应 

分 类 号：TN941.1[电子电信—信号与信息处理]