基于GAN的视景图像超分辨率重建方法研究  

Research on Super Resolution Reconstruction Method of Visual Image Based on GAN

作　　者：雷华成 刘伟 张云水 陈贤龙 黄胜[2] 

机构地区：[1]珠海翔翼航空技术有限公司,广东 珠海 [2]重庆邮电大学通信与信息工程学院,重庆

出　　处：《计算机科学与应用》2022年第12期2804-2812,共9页Computer Science and Application

摘　　要：飞行模拟器是一种用于复现飞行器及空中环境并能够进行模拟飞行操作的装置,在基于飞行模拟器系统进行日常飞行模拟训练时,视景系统所呈现出来的图像质量的好坏将直接影响最终的训练效果。图像超分辨率重建是一种能提升图像质量的有效手段。结合生成对抗网络(GAN)在视觉效果上有着很好的优越性,提出一种基于GAN的视景图像超分辨率重建方法,以提高飞行模拟视景图像显示效果。整个网络由生成器和鉴别器构成,其中生成器由浅层特征提取部分、深层特征提取部分和图像重建三个部分组成,鉴别器采用相对鉴别的思想实现对真伪图像的鉴别。在不同测试集上的实验结果表明,该方法不仅在视觉效果上优于RDN、DRFN等网络,同时,在客观评价指标上也好于它们。Flight simulator is a kind of simulation device which can reproduce aircraft and air environment and simulate flight operation. In the daily flight simulation training based on flight simulator system, the quality of the image presented by the visual system will directly affect the final training effect. Image super-resolution reconstruction is an effective method to improve image quality. Combined with the advantages of Generative Adversarial network (GAN) in visual ef-fects, a super resolution reconstruction method of visual image based on GAN is proposed to improve the visual image display effect of flight simulation. The entire network is composed of generator and discriminator, of which the generator is composed of three parts: shallow feature extraction part, deep feature extraction part and image reconstruction, and the discriminator adopts the idea of relative discrimination to achieve the identification of true and false images. Experimental results on different test sets show that the proposed method is not only superior to RDN, DRFN and other networks in terms of visual effects, but also better than them in objective evaluation indicators.

关 键 词：视景系统 生成对抗网络 视景图像 超分辨率重建 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]