ADMM for image restoration based on nonlocal simultaneous sparse Bayesian coding

Highlights：

• We propose a novel group-based simultaneous sparse Bayesian coding model (SSBC), which inherits the merit of group-based simultaneous sparse coding model and weighted sparse coding model.

• We estimate the sharing scaling variables by empirical Bayesian strategy. With the estimation of scaling variables, the structured sparse matrix can be efficient estimation by its posterior mode. Our method is more effective than the popular AM method which often need alternating iteration between scaling variable and structured sparse matrix.

• We apply the proposed model to solve image denoising and form a denoiser, which can be efficiently plugged into the alternating direction method of multipliers framework to solve more general image restoration problems, e.g., image denoising, inpainting, deblurring and single image super-resolution.

摘要

•We propose a novel group-based simultaneous sparse Bayesian coding model (SSBC), which inherits the merit of group-based simultaneous sparse coding model and weighted sparse coding model.•We estimate the sharing scaling variables by empirical Bayesian strategy. With the estimation of scaling variables, the structured sparse matrix can be efficient estimation by its posterior mode. Our method is more effective than the popular AM method which often need alternating iteration between scaling variable and structured sparse matrix.•We apply the proposed model to solve image denoising and form a denoiser, which can be efficiently plugged into the alternating direction method of multipliers framework to solve more general image restoration problems, e.g., image denoising, inpainting, deblurring and single image super-resolution.

论文评审过程：Received 3 December 2017, Revised 3 August 2018, Accepted 23 September 2018, Available online 4 October 2018, Version of Record 11 October 2018.