Research For Single Image Super-resolution Reconstruction Based On Sparse Representation

As a technology which no need to improve hardware devices.but can significantly improve image resolution.the image super-resolution reconstruction has become a hot spot of research in the field of image processing.Recently.the reconstruction algorithms based on image sparse representation, which learning a pair of over-complete dictiontionaries from example images and a sparse representation coefficient of the test image over the low-resoluton dictionary.have achived better reconstruction results.and this kind of algorithms posses a strong robust to noise.In order to further improve the reconstruction’s quality and the efficiency of dictionary learning.this paper focuses on the survey of image super-resolution reconstruction based on the sparse representation and the new dictionary learning methods.The main works are as follows:(1)In order to reduce the amounts of training samples, and thus improving the efficiency of dictionary learning without debasing the ability to representation of over-dictionary,a method of choose training images based image self-similarity is given.In this method.the single test image is treated as the high resolution training image.and the low resolution training image is down-sampled from a blurred version of the high resolution training image.(2)Under the premise of single test image is treated as the high resolutiontraining image.a image super-resolution reconstruction algorithm based on K-SVD(Sigular Value Decomposition) algorithm is proposed in this paper. The K-SVD algorithm can effectively reduce the number of atoms of the dictionary in the training, and the training atoms that can still linear express initial dictionary. Thereby effectively improving the sparse coding efficiency of the test image over the over-complete dictionary. The simulation results show that the algorithm does improve the efficiency of image reconstruction, but also greatly improve the quality of image reconstruction.(3)A novel dictionary learning methods named SCDL(Semi-coupled dictionary learning) is used to the image super-resolution reconstruction.The tradition methods assumed that exist coupled dictionaries of HR and LR images.which have the same sparse representation for each pair of HR and LR images.Under SCDL,the above assumption is relaxed and a pair of dictionaries and a mapping function are learned simultaneously. Simulation results show that this algorithm achives better image reconstruction quality than traditional sparse representation methods.