Image restoration using structured sparse representation with a novel parametric data-adaptive transformation matrix
作者：
Highlights：
• A novel framework for image restoration by combining nonlocal regularization technique and structured sparse estimation is proposed.
• A parametric nonlocal difference operator is proposed to give flexibility in modeling prior information of natural images.
• A novel parametric data-adaptive transformation matrix is introduced to construct sparse representation dictionary.
• The sparse nature of the transform coefficients of image patches under the proposed transformation matrix is exploited to restore degraded images.
• Neither nonlocal self-similarities in natural images nor patch clustering algorithms are required in our proposed structured sparse model.
摘要
Highlights•A novel framework for image restoration by combining nonlocal regularization technique and structured sparse estimation is proposed.•A parametric nonlocal difference operator is proposed to give flexibility in modeling prior information of natural images.•A novel parametric data-adaptive transformation matrix is introduced to construct sparse representation dictionary.•The sparse nature of the transform coefficients of image patches under the proposed transformation matrix is exploited to restore degraded images.•Neither nonlocal self-similarities in natural images nor patch clustering algorithms are required in our proposed structured sparse model.
论文关键词：Image restoration,Structured sparse model,Nonlocal difference operator,Nonlocal regularization,Transformation matrix
论文评审过程：Received 15 June 2016, Revised 2 January 2017, Accepted 6 January 2017, Available online 10 January 2017, Version of Record 23 January 2017.
论文官网地址：https://doi.org/10.1016/j.image.2017.01.003
