Research On Super-resolution Reconstruction Algorithm Based On Residual Dictionary Learning

With the rise of artificial intelligence and machine learning,image super-resolution reconstruction has become one of the most popular research directions in the field of image processing and computer vision.High-resolution image reconstruction technology has important value in many application fields,and provides theoretical and technical support for in-deep fields such as image recognition,computer vision and artificial intelligence.In order to improve the resolution of the reduced image,the image super-resolution reconstruction algorithm based on the residual dictionary is studied and improved in this paper.1.Super-resolution algorithm based on residual dictionary image.Firstly,the fuzzy matrix and down-sampling technique are used to degrade the high-resolution image set to obtain the corresponding low-resolution image set.The high-resolution residual image set is obtained by comparing the high-resolution image set with the low-resolution image set which is reconstructed by the Bicubic interpolation.The residual image contains the high frequency information.Secondly,the sparse residual dictionary pair are obtained by training joint dictionary through the residual maps which are training sample.Finally,the image super-resolution reconstruction is obtained by the high-resolution residual dictionary and the sparse coefficient of the degraded image which is calculating through the low-resolution residual dictionary.The experimental results show that the proposed algorithm performs well in both subjective and objective evaluation of reconstructed images.2.Super-resolution reconstruction algorithm based on image fusion,The high frequency information in the low resolution image also plays animportant role in the high frequency information of the super-resolved image,so extracting the high frequency part of the low-resolution image block has a key influence on the training of the sparse dictionary and solving the sparse coefficient,and the image can be reconstructed better.Therefore,the texture and edge features of the extracted image are used as the sample set of the training sparse dictionary,so that the texture dictionary and the edge dictionary are trained;two pseudo-high resolution images are reconstructed by two feature sparse dictionaries;finally,two images are realized by wavelet transform of DTW.The reconstructed image is optimized and processed by iterative inverse projection.The experimental results show that the algorithm improves the image reconstruction to a certain extent,highlights the edge and texture information of the image,and provides us with more detailed image information.