Image Super-Resolution Based on Structure-Modulated Sparse Representation

   December 11, 2017    Comments Off on Image Super-Resolution Based on Structure-Modulated Sparse Representation    Posted in: IEEE 2017, MATLAB

Image Super-Resolution Based on Structure-Modulated Sparse Representation

ABSTRACT:

Sparse representation has recently attracted enormous interests in the field of image restoration. The conventional sparsity-based methods enforce sparse coding on small image patches with certain constraints. However, they neglected the characteristics of image structures both within the same scale and across the different scales for the image sparse representation. This drawback limits the modeling capability of sparsity-based super-resolution methods, especially for the recovery of the observed low-resolution images. In this paper, we propose a joint super-resolution framework of structure modulated sparse representations to improve the performance of sparsity-based image super-resolution. The proposed algorithm formulates the constrained optimization problem for high resolution image recovery. The multistep magnification scheme with the ridge regression is first used to exploit the multi-scale redundancy for the initial estimation of the high-resolution image. Then, the gradient histogram preservation is incorporated as a regularization term in sparse modeling of the image super resolution problem. Finally, the numerical solution is provided to solve the super-resolution problem of model parameter estimation and sparse representation. Extensive experiments on image super-resolution are carried out to validate the generality, effectiveness, and robustness of the proposed algorithm. Experimental results demonstrate that our proposed algorithm, which can recover more fine structures and details from an input low-resolution image, outperforms the state-of-the-art methods both subjectively and objectively in most cases.

 

EXISTING SYSTEM:

  • In real world scenarios, the low-resolution (LR) images are generally captured in many imaging applications, such as surveillance video, consumer photographs remote sensing, magnetic resonance (MR) imaging and video standard conversion.
  • The resolution of images is limited by the image acquisition devices, the optics, the hardware storage and other constraints in digital imaging systems. However, high-resolution (HR) images or videos are usually desired for subsequent image processing and analysis in most real applications. As an effective way to solve this problem, super-resolution (SR) techniques aim to reconstruct HR images from the observed LR images.
  • The super-resolution reconstruction increases high-frequency components and removes the undesirable effects, e.g., the resolution degradation, blur and noise. Recently, numerous SR methods have appeared to estimate the relationship between the LR and HR image patches with promising results. Some typical methods usually need a large and representative database of the LR and HR image pairs.

DISADVANTAGES OF EXISTING SYSTEM:

  • Existing system increase edge halos, blurring and aliasing artifacts.

PROPOSED SYSTEM:

  • We propose a novel joint framework of the structure-modulated sparse representation (SMSR) for single image super-resolution. The multi-scale similarity redundancy is investigated and exploited for the initial estimation of the target HR image. The image gradient histogram of a LR input is incorporated as a gradient regularization term of the image sparse representation model. The proposed SMSR algorithm employs the gradient prior and non locally centralized sparsity to design the constrained optimization problem for dictionary training and HR image reconstruction. The main contributions of our work can be summarized as follows:
  • The multi-step magnification scheme with the ridge regression is proposed to initialize the target HR image for the solution of image SR problem;
  • The novel sparsity-based super-resolution model is proposed with the combination of multiple image priors on the structural self-similarity, the gradient histogram and the nonlocal sparsity;
  • The gradient histogram preservation (GHP) is theoret-ically deduced for image SR reconstruction and also incorporated as the regularization term for the sparse modeling of HR image recovery.

 

ADVANTAGES OF PROPOSED SYSTEM:

  • Our proposed algorithm can recover more fine structures and details from an input low-resolution image, outperforms the state-of-the-art methods both subjectively and objectively in most cases.

 

SYSTEM REQUIREMENTS:

HARDWARE REQUIREMENTS:

 

  • System : Pentium IV 2.4 GHz.
  • Hard Disk : 40 GB.
  • Floppy Drive : 44 Mb.
  • Monitor : 15 VGA Colour.
  • Mouse :
  • Ram : 512 Mb.

 

SOFTWARE REQUIREMENTS:

 

  • Operating system : Windows XP/7.
  • Coding Language : MATLAB
  • Tool : MATLAB R2013A

REFERENCE:

Yongqin Zhang, Member, IEEE, Jiaying Liu, Member, IEEE, Wenhan Yang, and Zongming Guo, Member, IEEE, “Image Super-Resolution Based on Structure-Modulated Sparse Representation”, IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 24, NO. 9, SEPTEMBER 2015.

 

  

Comments are closed.