Abstract:Row nonlocal similarity regularization term with l1-norm constraint is explored.Dual-sparsity regularized sparse representation model is presented for SISR.Iterative shrinkage algorithm is extended to solve dual l1-norm constraints model. Recently, by exploring the column nonlocal similarity prior among the sparse representation coefficients, the column nonlocal similarity sparse representation models for solving the ill-posed single image super-resolution (SISR) problem are attracting more and more attention. However, these conventional models consider only the prior among nonlocal similar sparse representation coefficients, and fail to consider the prior among all entries (or rows) of the sparse representation coefficient. Hence the modeling capability may be limited. In fact, if a cluster of similar representation coefficients is rearranged into a matrix in the sparse representation coefficient space, the nonlocal similarity priors exist both among columns and rows. Using the row nonlocal similarity prior, a row nonlocal similarity regularization term with l1-norm constraint is explored. By introducing it to the conventional column nonlocal similarity sparse representation model, we present a dual-sparsity regularized sparse representation (DSRSR) model. A surrogate function based iterative shrinkage algorithm is introduced to effectively solve the proposed model. Extensive experiments on SISR demonstrate that the presented model can effectively reconstruct the edge structures and suppress the noise, achieving convincing improvement over many state-of-the-art example-based methods in terms of PSNR, SSIM and visual quality.

Dual-sparsity regularized sparse representation for single image super-resolution.

Sparse Representation Based Single Image Super-Resolution with Low-Rank Constraint and Nonlocal Self-Similarity

Multiscale Self-Similarity And Sparse Representation Based Single Image Super-Resolution

Bidirectionally Aligned Sparse Representation for Single Image Super-Resolution

Compressive Sampling based Single-Image Super-resolution Reconstruction by dual-sparsity and Non-local Similarity Regularizer

Single Image Super Resolution Using Local Smoothness and Nonlocal Self-Similarity Priors

Single Image Super-Resolution Via Adaptive Transform-Based Nonlocal Self-Similarity Modeling and Learning-Based Gradient Regularization

Single Image Super-Resolution Via Adaptive Sparse Representation and Low-Rank Constraint.

Single Image Super-Resolution Based on Nonlocal Sparse and Low-Rank Regularization.

Kernel based low-rank sparse model for single image super-resolution

Image Super-Resolution by Learning Weighted Convolutional Sparse Coding.

Single Image Superresolution Using Maximizing Self-Similarity Prior

Subspace Constraint for Single Image Super-Resolution

Image Super-resolution Based on Sparse Representation With Joint Constraints

Single Image Super-Resolution by Learned Double Sparsity Dictionaries Combining Bootstrapping Method

Super-resolution with Nonlocal Regularized Sparse Representation

Image Super-Resolution Reconstruction Based on L1/2 Sparsity

Low-rank Sparse Representation for Single Image Super-Resolution Via Self-Similarity Learning

Single Image Super-Resolution Based on Nonlocal Similarity and Sparse Representation

Image Super-Resolution Via RL-CSC: when Residual Learning Meets Convolutional Sparse Coding

Boosting Single Image Super-Resolution Learnt from Implicit Multi-Image Prior.