Research On Single Image Super-Resolution Based On Sparse Representation

Image super-resolution technology has become a pivotal research in image processing.It has been widely used in many fields such as remote sensing,surveillance security and medical science.Using SR technology to enhance the resolution of image or video can give people a good visual enjoyment,and conducive to the identification and analysis of the target in the image.This paper will study the super-resolution reconstruction algorithm via sparse representation,and analyze topics such as dictionary design,feature extraction and quality evaluation.The main research contents and conclusions include the following:This paper will analyze the mathematic principle of sparse representation and the algorithm flow based on the observation model,including the key issues such as the design of overcomplete dictionary,sparse decomposition algorithm and regularization optimization.In addition,several other super-resolution reconstruction algorithms are briefly introduced,e.g.interpolation,reconstruction,and machine learning.The advantages and disadvantages of each algorithm are compared finally.This paper improves the defect that the original high-frequency details are not sufficiently reconstructed.We divide the high-frequency components into two parts: the main high-frequency component and the residual high-frequency component.The high-frequency information is trained in the main high-frequency dictionary pairs.By subtracting the estimation from the original,we obtain residual mid and high frequency images,which is used to training the residual high-frequency dictionary pairs.Finally we use the double dictionary pairs to reconstruct low-resolution images.Compared with the original algorithm,the algorithm proposed in this paper has obvious improvement in both subjective and objective evaluation,and can reconstruct more texture and detail information.So the reconstructed image is more closer to the original image.This paper develops a relatively complete solution of evaluation programs based on many evaluation experiments.Using the combination of subjective and objective evaluations to standardize the content and process of evaluation.The subjective and objective indicators we select can reflect the overall effect and local details of image,therefore it improves the reliability and standardization of quality evaluation.