Research Of Super-Resolution Image Reconstruction Technique Based On Solo Dictionary For Sparse Representation

Image super-resolution reconstruction is a technique that reconstructs a high-resolution image of rich details from single or multiple low resolution images of a same scene.The technique overcomes the limitation of the imaging system,improving the image resolution economically and effectively.In the super resolution reconstruction technology,sparse representation based image super-resolution reconstruction technique is a hot topic in the field of image processing because of its widely adaptability and good reconstruction result.However,due to the fixed image patch size in local processing,there is a limit of accuracy in matching from low to high resolution image patch.By the other hand,there is a flaw that the dictionary needs to be retrained when the zooming factor changes.In this paper we propose some methods to solve above-mentioned problem:(1)Get better local matching by using the adaptive method in adjusting size of image patch.Take advantage of high and low frequency existed in the image so that the image can be divided into smooth regions and edge regions.Then reconstruct image patch by using adaptive size of image patch distributed in different regions.(2)Replace joint over-complete dictionary learning with solo over-complete dictionary learning.Regard low-resolution image patch as missing pixels of corresponding high-resolution image patch.In this way,we propose a novel solution for local processing which makes the dictionary need not to be retrained when the zooming factor changes.We did the full experiment to prove the validity of the method.And the experimental results show that the proposed method has better performance in both subjective and objective image quality evaluation.