Research On Super Resolution Image Restoration Based On Redundant Dictionary

High resolution images not only meet the needs of scientific research applications,but also enable users to experience a good sense.However,it is very difficult to obtain high resolution images because of the constraint of image acquisition equipment.So how to reconstruct the low resolution image and get the desired high resolution image is a very important task,and has a wide range of applications.The difficulty of image super-resolution reconstruction is that the observed low resolution image contains not enough information to restore high resolution images,which is also a mathematical problem.In the extensive application of super-resolution image reconstruction algorithm,super-resolution image restoration algorithm based on sparse representation is widely used.The sparse representation of the super-resolution reconstruction algorithm to low resolution image blocks based on the training,then the image block training into high and low resolution dictionary,will test the image map to get low resolution dictionary sparse coefficients,then multiplied by the high-resolution dictionary to obtain the final reconstructed image.The main research contents are as follows:Two of the most important parts of the sparse representation theory are sparse coefficients and dictionary learning,this paper focuses on solving the sparse coefficient and redundant dictionary learning,use different sparse coefficient algorithm and redundant dictionary learning for super resolution image restoration obtained high resolution image restoration.Training dictionary in super-resolution image restoration based on dual redundant global dictionary learning using K-SVD algorithm.In the recovery process,the training dictionary is used to restore the high resolution high frequency and high resolution residual high-frequency,and the two restoration results are added to get the final restoration image.The second dictionary trainingstructure blocks the most similar matching blocks of the image blocks to form a structural group.Super resolution image restoration based on PCA_KLLD redundant local dictionary.Local dictionary learning algorithm is used for image reconstruction,and two local dictionary learning and global dictionary learning are compared by simulation experiments.The defects of super resolution image restoration DUAL_KLLD local Dictionary of high resolution image restoration are based on previous(block effect)was improved,using the improved PCA_KLLD local multi redundant dictionary learning complex original image super resolution image restoration.Two redundant multi local dictionary based super-resolution image restoration and K-SVD global dictionary super-resolution image restoration are compared with the simulation experiments,and the experimental data are used to draw conclusions.