Image Super Resolution Reconstruction Based On Edge Enhancement

Super resolution(SR)image reconstruction is to recover high resolution(HR)image with clear edge texture and rich details from observed low resolution(LR)image.In reality,due to the limitation of hardware conditions,improving the image quality will lead to the increase of cost,so the research of image SR reconstruction technology is particularly important.Image SR reconstruction technology has important application value in security monitoring system,medical imaging,military remote sensing,satellite imaging and other fields.At present,image SR reconstruction methods based on traditional methods have achieved good results,but there is still a problem that the image edge is blurred due to insufficient reconstruction of texture details.Therefore,this paper focuses on the problem of high-frequency detail loss in current traditional image reconstruction,and two image SR methods based on edge enhancement are proposed.1.An image SR reconstruction method based on gradient regularized edge enhancement is proposed.Firstly,principal component analysis is used to learn many kinds of concise principal component dictionaries,and the sparse region is adaptively selected by judging the distance between the target image block and the center of the dictionary class.Then,in order to reconstruct more image details,the neighborhood regression method is used to learn the gradient prior information of the image from the external examples,and an image gradient regularization term is constructed to guide the edge detection of the image.In order to improve the quality of the reconstructed image,edge detail reconstruction is used to make up for the lack of prior image.Experimental results show that the gradient regularized edge enhanced image SR reconstruction algorithm proposed in this paper obtains better reconstruction results,is more robust to noise,and can effectively reconstruct the lost details of the image.The subjective and objective reconstruction results are better than most methods.2.An image SR reconstruction method based on random forest edge enhancement is proposed.In order to improve the quality of the reconstructed image,firstly,the input LR image is directly enlarged to the same size as the image to be reconstructed by bicubic interpolation method to realize the low-frequency reconstruction of the image.Then,the gradient blocks are extracted from the input LR image,and the random forest is used for classification,and the regression model is introduced to learn the mapping relationship between the high frequency blocks of LR image and the high frequency blocks of HR image,so as to realize the reconstruction of the high frequency components of the image.Finally,the estimated HR image is obtained by minimizing the energy function including the low frequency consistency and high frequency adaptability,and the obtained HR image is of high quality The reconstructed highfrequency components are included,so the edge is enhanced.The experimental results show that the SR image reconstruction method based on random forest edge enhancement proposed in this paper obtains better reconstruction results,and can restore clearer texture and edge information.The subjective and objective reconstruction indexes are better than most methods.