Research On Super-resolution Reconstruction Algorithm Of Cultural Relic Images

Digital protection and classification of cultural relics is one of the current research hotspots.Due to the limitation of uncontrollable factors such as objective conditions,the resolution of the obtained cultural relic images is often affected in the real environment,hindering the development of cultural relics digital processing.The super-resolution technology of cultural relic images improves the resolution of cultural relic images and enhance the visual effects.It has important research value in the field of cultural relics digitization.After researching and analyzing the basic generative adversarial network and conventional algorithms,this article proposes an improved super-resolution algorithm for cultural relic images.The main tasks of this article are as follows:（1）This article first introduces some theoretical foundations in the field of image super-resolution and the direction of deep learning.At the same time,it analyzes the principles and defects of basic generative adversarial networks from the perspective of mathematical theory derivation.On this basis,this article conducts experimental analysis on a variety of conventional image super-resolution models,determines the image super-resolution algorithm SPSR guided by gradient information as the basic model.（2）This article proposes IGAN,a super-resolution algorithm for cultural relic images based on the strategy of information block extraction.The IGAN algorithm is based on the SPSR model and establishes two cultural relic image data sets,RCdata and FCdata.In order to combine the super-resolution model with the cultural relic images,this method first considers the problem of low utilization of high-frequency information in the original high-resolution cultural relic images,explores an improved information block extraction Strategy;then consider the advantages and disadvantages of different up-sampling methods,replace the original nearest neighbor interpolation with sub-pixel convolution.The experimental results show the applicability and superiority of the IGAN algorithm on cultural relic images.（3）This article proposes a super-resolution algorithm CPIGAN for cultural relic images based on the pyramid structure of the closed-loop.In view of the shortcomings of the IGAN algorithm,the CPIGAN algorithm starts from the internal architecture.First,it designs a generation network based on a pyramid structure to improve the ability of the network model to extract the characteristic information of cultural relic images.Then,the closed-loop structure is incorporated into the generation network to make the model better learn the mapping between low-high resolution cultural relic images.Through experiments,compared with different algorithms,the image of cultural relics reconstructed by the CPIGAN algorithm is of higher quality and the texture details are clearer.