Three-dimensional fractional total variation regularized tensor optimized model for image deblurring
作者：
Highlights：
• We propose a tensor-based FTV model for the three-dimensional image deblurring problem.
• The tensor algebra facilities 3DFTV and blurring operator directly on images instead of transforming them into vector forms, which retains the intrinsic structure in images.
• Our model can sufficiently recover texture and alleviate staircase effects by exploiting the nonlocal smoothness of images compared with some classical TV-based models.
• We use the ADMM-based algorithm to solve our 3DFTV-based model. This algorithm splits our model-based problem into variable subproblems. Each subproblem has an easier closed-form solution, which makes the algorithm efficiently implementable.
摘要
•We propose a tensor-based FTV model for the three-dimensional image deblurring problem.•The tensor algebra facilities 3DFTV and blurring operator directly on images instead of transforming them into vector forms, which retains the intrinsic structure in images.•Our model can sufficiently recover texture and alleviate staircase effects by exploiting the nonlocal smoothness of images compared with some classical TV-based models.•We use the ADMM-based algorithm to solve our 3DFTV-based model. This algorithm splits our model-based problem into variable subproblems. Each subproblem has an easier closed-form solution, which makes the algorithm efficiently implementable.
论文关键词：Fractional total variation,Tensor,Staircase effects,Image deblurring,Three-dimensional images
论文评审过程：Received 20 August 2019, Revised 18 March 2021, Accepted 24 March 2021, Available online 5 April 2021, Version of Record 5 April 2021.
论文官网地址：https://doi.org/10.1016/j.amc.2021.126224
