Research On Image Super-resolution Reconstruction Based On Sparse Representation

High-resolution(HR)image contains a lot of detail information,which benefits for diagnosis,analysis and further processing of the image,so it is crucial for medical imaging,video monitoring,remote sensing and other fields.Image super-resolution(SR)reconstruction is a technique that uses the image processing method to restore a HR image from one or more low-resolution(LR)images of the same scene,and it can improve the resolution of the image quickly and low costly without change the original system hardware.At present,SR has become a research focus in the field of image processing,computer vision,pattern recognition and so on.This paper had comparatively done systematic study to the application of sparse representation in image SR reconstruction.And the existed algorithm based on sparse representation is improved,which is applied to the super resolution reconstruction of medical image.On the basis of traditional image SR reconstruction based on the sparse representation,the selection of dictionary samples and reconstruction process are improved.In order to make the dictionary learn more high frequency information,the combined information of the Pyramid upper images generated by self-similarity and the natural images database are used.In addition,in the process of reconstruction,there are two important aspects.Firstly,the Pyramid top image is taken as the initial estimation to improve the reconstruction effect.Secondly,according to the non-local structure self-similarity of image,the information contained in the same scale similar image patches,which is used to be the regularization constraints,is added into the reconstructed image to constrain and modify the reconstruction result.Simulation experiments were carried out on natural images and medical images,and the experimental results show the proposed method is effective for improving the effect of image SR reconstruction.