基于深度学习的显著运动下高动态范围图像重建的研究  

Research on High Dynamic Range Image Reconstruction under Significant Motion Based on Deep Learning

作　　者：黄天鑫 姜竹青[1] 贾梦珍 郑栀芯 Huang Tianxin;Jiang Zhuqing;Jia Mengzhen;Zheng Zhixin(Beijing University of Posts and Telecommunications,Beijing 100866,China)

机构地区：[1]北京邮电大学,北京100866

出　　处：《广播与电视技术》2024年第7期30-33,共4页Radio & TV Broadcast Engineering

基　　金：基金会项目(No.A14B07C02-202305D1)资助。

摘　　要：高清直播业务在广电行业中举足轻重,但是存在不同强度光照以及人物运动导致直播画面不清晰和亮度不自然的问题,本文提出了一种基于深度学习技术的方法,可应用于显著运动下的高动态范围图像重建。实验结果表明,该方法不仅改善了画面清晰度,还解决了亮度不均和运动模糊等问题,进一步提升了用户的直播观看体验。此外,该方法还大幅降低了模型计算成本,为智能高清直播技术的进一步推进提供了有力支持。The high-definition live broadcasting service is crucial in the broadcasting industry.However,challenges arise due to varying intensity of lighting and the movement of subjects,leading to issues such as unclear images and unnatural brightness in live broadcasts.This paper proposes a method based on deep learning technology for reconstructing high dynamic range images under significant motion.Experimental results demonstrate that this method not only enhances image clarity but also addresses issues like uneven brightness and motion blur,thereby further improving the user's live broadcasting viewing experience.Additionally,the proposed method significantly reduces the computational cost of the model,providing robust support for the advancement of intelligent high-definition live broadcasting technology.

关 键 词：高动态范围图像重建 深度学习 显著运动图像 HSV颜色空间 

分 类 号：TP319[自动化与计算机技术—计算机软件与理论]