摘要
激光图像在采集和传输过程中,易受到其它因素的影响,会出现破损现象,针对传统技术无法实现高精度激光图像修复的难题,为此设计了基于深度神经网络的激光图像修复技术。首先对当前激光图像修复的研究现状进行分析,找到引入激光图像修复精度低的原因,然后采用深度学习神经网络对激光图像受损区的轮廓线结构进行映射,根据利用曲线拟合理论将激光图像受损区分割出来,最后对激光图像受损区的相关信息进行填充,完成激光图像修复,并进行了激光图像修复的仿真测试。结果表明,相对于原始激光图像,修改后的激光图像不仅信噪比大幅度提升,激光图像的视觉效果更佳,而且激光图像修复效果要明显优于传统激光图像修复技术,验证了本文激光图像修复技术的优越性。
In the process of laser image acquisition and transmission,it is easy to be affected by other factors,and there will be damage.Aiming at the problem that traditional technology can not achieve high-precision laser image res-toration,a laser image restoration technology based on depth neural network is designed.Firstly,the current research status of laser image restoration is analyzed,and the reasons for the low precision of laser image restoration are found.Then,the contour structure of the damaged area of laser image is mapped by deep learning neural network.According to curve fitting theory,the damaged area of laser image is separated.Finally,the relevant information of the damaged area of laser image is filled to complete laser image restoration.The simulation results show that compared with the original laser image,the signal-to-noise ratio of the modified laser image is greatly improved,the visual effect of the laser image is better,and the laser image restoration effect is obviously better than the traditional laser image restoration technology,which verifies the superiority of the laser image restoration technology in this paper.
作者
姚晓峰
须文波
武利秀
YAO Xiaofeng;XU Wenbo;WU Lixiu(Jiangsu Key Construction Laboratory of IoT Application Technology,Taihu University of Wuxi,Wuxi Jiangsu 214046,China;School of Internet of Things Engineering,Jiangnan University,Wuxi Jiangsu 214122,China)
出处
《激光杂志》
北大核心
2019年第11期76-79,共4页
Laser Journal
基金
教育部中国移动科研基金项目(No.MCM20170204)
关键词
深度神经网络
激光图像
清晰度
修复特征
图像信噪比
depth neural network
laser image
sharpness
repair features
image signal-to-noise ratio
