The Research On Text Image Super-resolution Reconstruction Method Based On Sparse Representation

Along with the rapid development of information technology,information processing technology is required to improve.High resolution images or video is usually required in most digital image applications and image processing and analysis.Nowadays,the hardware cost is not a problem,but for most of the low resolution text images that have been defaced,even if the hardware is clear enough,the text is not clear,in this case,text image super resolution is particularly important.Many scholars at home and abroad have done a lot of researches on super-resolution reconstruction,and their algorithms have been successfully applied to natural images but ineffectively in text images.Text image is a kind of unique image,so it is necessary to study the specific technology which is suitable for it.Some scholars have put forward some algorithms for text images,but there are two problems,one is the high complexity of the algorithm,and the other is not good reconstruction result when pirior information is insufficient.Therefor,based on the characteristics of the text image,the sparse representation super-resolution method is studied in this paper.We have improved the algorithm from efficiency and accuracy.The specific research contents are as follows:(1)We have studied the degradation model of the text image,and have analyzed several current image reconstruction models and dictionary training algorithm,and the specific process of the sparse representation reconstructionalgorithm has also been studied and analyzed.(2)We have analyzed the joint training method in sparse representation algorithm.In order to solve the problem of running for a long time and low efficiency,we proposed a method to optimize the dictionary training.Only high resolution dictionary is learned from trainning sample,but the low resolution dictionary is derived by HR dictionary in the method.In this way,the operation time is shortened and the efficiency is improved.In the high resolution dictionary learning stage,the K-SVD algorithm has been used to train the HR dictionary.In the sparse coefficient solving phase,we have employed feature-sign algorithm to solve the convex problem after analysing the sparse coefficient model.Finally,compared with the existing algorithm from the reconstruction results and the execution time and the experimental results proves that the running time of our algorithm is shortened by 45.7%,at the same time the PSNR and SSIM values are slightly higher than the original sparse reconstruction algorithm.Our algorithm ensures the accuracy while improving the execution efficiency.(3)The reconstruction images of original sparse representation method is not clear,and distinguish between foreground and background is not obvious,and the edges are not continuouse.Therefor we have studied the characteristics of the text image on the basis of optimizing the dictionary training method,then the bimodal restriction characteristic of the introduced text image is constrained to reconstruct the high resolution image as a regular term,and edge optimization algorithm is used to optimize the edge of the image.The simulation and analysis of the proposed algorithm are carried out and compared with original sparse reconstruction method.The results show that the proposed algorithm has better performance and higher precision.(4)We have analyzed the characteristics of the inscription image,and have designed the image processing flow according to the characteristics of the inscription image.The inscription image is preprocessed,and then the image isrestored by the improved reconstruction algorithm.At last,the feasibility and practicability of the algorithm are proved by experiments.The experimental results show that the edge of the image is clear,the text background is clear,and the text is easy to identify.