摘要
针对Tikhonov正则化的预处理共轭梯度图像复原算法中模糊图像取全零扩展矩阵的不足之处,研究了零边界条件下Tikhonov正则化的预处理共轭梯度算法.提出了新的模糊图像的扩展矩阵,降低了原矩阵向量积的计算误差,修正了初始梯度的取值.改进算法更符合真实的图像退化过程,有效提高了复原的图像质量.实验结果表明:对于各种退化造成的模糊图像,与当前求解全变分正则化的IST、TwIST、SALSA算法比较,本文算法复原效果优于当前流行的图像复原算法.
To deal with the shortcoming of the preconditioning conjugate gradient （PCG） method with Tikhonov regularization in which the blurred image is extended with a zeros extension matrix. With the careful analysis of the PCG method with Tikhonov regularization under zero boundary condition, a new extension matrix for the blurred image was proposed. It could decrease the matrix vector multiplication computational error and modify the initial gradient. The improved algorithm is in accord with the real image blurring process and increases the quality of recovered image. Experiments show that, compared with the state-of-the-art algorithms of IST, TWIST and SALSA which solve the total-variation （TV） regularization, the proposed algorithm performs favorably.
出处
《北京理工大学学报》
EI
CAS
CSCD
北大核心
2013年第9期980-984,990,共6页
Transactions of Beijing Institute of Technology
基金
国家"九七三"计划项目(2009CB72400603)
国家自然科学科学仪器专项资助项目(61027002)
国家自然科学基金资助项目(60972100)
