Research On Image Super-Resolution Based On Correlation Analysis Theory

The super-resolution method,as an emerging technology to improve the image resolution,has gradually become a hot field for people to study.The image super-resolution reconstruction method based on the canonical correlation analysis is taken into account that the canonical correlation analysis can maximize the correlation between the two variables,and the topology of the low-resolution image subspace is applied to the high resolution rate of image,and then high-resolution images can be restored.Experiments show that the image super-resolution reconstruction method based on canonical correlation analysis can clearly restore high-resolution images.In this paper,how to improve the effect of super-resolution reconstruction method based on the canonical correlation analysis is further studied.1.Presented an image super-resolution method based on supervised canonical correlation analysis.In this paper,We analysis and use three kinds of supervised canonical correlation analysis algorithms which is generalized canonical correlation analysis,discriminative correlation analysis and local discriminative canonical correlation analysis.The experimental results show that the effect of supervised canonical correlation analysis algorithm is betterr than that of unsupervised canonical correlation analysis.The model framework of image super-resolution method based on supervised canonical correlation analysis is given,and the experimental comparison is carried out.Indicating that a canonical correlation analysis with label information can better restore high-resolution images.2.A sparse reconstruction method of super-resolution based on discriminative canonical correlation is proposed.Because the neighbor values in the existing image super-resolution framework need to be set manually,so a new framework of super-resolution sparse reconstruction is given.When the CCA subspace is reconstructed,the sparse method is introduced into the model,and the optimal algorithm is used to select the nearest neighbor sample by using sparse selection of neighboring parameters without parameters.The efficiency of the method is improved and the better image can be restored.3.Proposed an image super-resolution sparse reconstruction method based on multi-view canonical correlation analysis.Because the existing canonical correlation analysis method only considers the information of a kind of view,it can not make full use of other visual angle information after using one class information.In this paper,a multi-view canonical correlation analysis method is proposed and applied to the image super-resolution reconstruction model.The experimental results show that the muti-view canonical correlation analysis can fully explore all kinds of view data based on the categories of information,information is minimized between class,and information within the class is maximized at the same time,the algorithm has a good recovery effect.