Research On Single Color Image Super-Resolution Reconstruction Based On Sparse Representation

The color image super-resolution reconstruction is a technique for reconstructing low-resolution color images into higher-resolution color images with larger size and better visual effects.Such technique have good application prospects in the field of high-definition digital television,satellite remote sensing monitoring,mobile internet,medical image processing,cultural relic protection and display.The learning-based method is one of the key directions of current color image super-resolution reconstruction research.Among them,the color image superresolution reconstruction method based on the sparse regular model has been widely used due to its good color information representation and relatively low computational complexity.The color image super-resolution reconstruction method based on sparse regular model,this paper takes natural images and mural images as objects,and improves some shortcomings in traditional sparse models.And the corresponding solution algorithm is proposed,as follows:1.Color Image Super-resolution Reconstruction Based on Color Constraint and Non-locally Sparse Representation algorithm is proposed.The sparse model of the reconstruction unit with image patch as the basic ignores the similarity among the nonlocal image patchs,which leads to problem that is easy to produce sparse representation error.Secondly,the super-resolution reconstruction of separated channel color images only reconstruct the luminance channel information,while the chrominance channel is processed by simple interpolation,ignoring the correlation among color channels,which may cause some color artifacts in color reconstructed image s.This paper proposes a color image super-resolution reconstruction algorithm that reconstructs high-resolution color images using non-local self-similarity sparse models and then reconstructs the correlation of color image channels by color channel constrained a priori enhancement.The algorithm can make full use of the color channel redundancy information of the reconstructed color image,thereby realizing effective superresolution reconstruction of the color images.Through the experiments on natural color images,it can be seen that the proposed algorithm has better performance in reconstructing detail of color image and color recovery.2.A super-resolution reconstruction algorithm for color mural image based on weighted kernel norm sparse regular model is proposed.In a color mural image,since there are a large number of non-local self-similar image blocks having similar structures and textures,etc.,the structural sparse matrix of the color mural image has a low rank characteristic.This paper proposes a color mural image super-resolution reconstruction algorithm based on the weighted kernel norm sparse regularizatio n model.The algorithm proposes a reliable sample construction principle for co lor mural images,which provide an effective sample set for sparse dictionary training.Then,the low-rank optimization regularity of the color mural image sparse matrix is weig hted by the weighted kernel norm,which effectively improves the super-resolution reconstruction quality of the color mural image.Experiments show that the proposed algorithm can improve the effect of super-resolution reconstruction of color mural images and improve the color fidelity of reconstructed color mural images.