As the key technology of infrared image resolution enhancement,single image super-resolution reconstruction algorithm has been widely studied in recent years.In order to improve the resolution of infrared image super-resolution reconstruction,a method of single infrared image super-resolution reconstruction based on dense residual generation network is proposed.Different from the previous super-resolution methods based on the generative countermeasure network,the novelty of this method mainly includes two aspects.Firstly,improve the network architecture to improve performance.The dense residual network is designed as the generation network of the generative countermeasure network,which makes full use of the hierarchical characteristics of the low resolution image.A continuous memory mechanism is introduced in the generation network to make use of the dense remaining blocks.Secondly,Wasserstein GAN is used as the loss function to modify the discriminant network model to achieve the purpose of stable training.A large number of experiments have been carried out by using infrared high resolution image data set.The results show that the method is superior to the latest method in both objective and subjective evaluation.
参考文献
相似文献
引证文献
引用本文
贾宇,温习,王晨晟.密集残差网络红外图像超分辨率重构[J].激光与红外,2020,50(10):1283~1288 JIA Yu, WEN Xi, WANG Chen-sheng. Super resolution reconstruction method with dense residual network for infrared image[J]. LASER & INFRARED,2020,50(10):1283~1288