Research On Image Super-Resolution Reconstruction Technology Based On Sparse Representation

To improve the resolution of the image is always an important topic in the field of image processing, due to the restrictions of imaging-forming system and technology, improving image resolution is difficult and with high cost in hardware aspect, so it is significant to achieve image super-resolution reconstruction by software method. Image super resolution(SR) technology aims to recover a clear high resolution image from one or more low-resolution images in same scene, namely to reconstruct those information that many factors result in lost in the image-getting process. At present there are many methods of image super-resolution reconstruction technology. Firstly, the methodes based on sparse theory have widely application in image compression, egmentation, restoration, noise an so on; Secondly, compressive sensing theory shows that natural images are sparse, which makes that based on sparse theoretical method to achieve image super resolution possible.The main research of this paper focuses on research and analysis of sparse theory, dictionary learning and cartoon-texture image decomposition algorithm, and the paper proposes a improved super-resolution algorithm to achieve a perfect vision effect and a high score in PSNR. The main research contents of this paper are:1. This paper introduces the research background and realistic significance of image super-resolution technology, current situation and development trend of SR, and now the existing problems and new solutions of image super resolution tech-nology;2. The sparse theory, sparse optimization algorithm and application of sparse theory are described systematically in this paper. This paper introduces some major process, including classic model of the algorithm of super resolution based on sparse representation, dictionary learning and image reconstruction. A novel algorithm that reconstruct a high resolution image by single dictionary learning based on the classic model is presented. The efficiency of reconstructed image is improved, as well as the computational complexity of dictionary learning is reduced.3. Related to cartoon-texture decomposition algorithm, a super resolution algorithm based on cartoon-texture decomposition and sparse representation is proposed. Finally, the experimental results show that the proposed algorithm has better performance.