面向强光源视频的帧插值方法  

Frame interpolation method for video with high light source

作　　者：辛伊宁 刘勇[1] 许骞艺 XIN Yi-Ning;LIU Yong*;XU Qian-Yi(College of Software Engineering,Heilongjiang University,Harbin 150080,China)

机构地区：[1]黑龙江大学软件工程学院,哈尔滨150080

出　　处：《黑龙江大学工程学报》2023年第2期57-64,共8页Journal of Engineering of Heilongjiang University

基　　金：国家自然科学基金项目(61972135,61602159);黑龙江省自然科学基金项目(LH2020F043,F201430);哈尔滨科技创新人才研究专项资金项目(2017RAQXJ094)。

摘　　要：视频帧插值是在原始帧之间合成不存在的帧,以提高视频帧率。常见的帧插值方法基于深度感知进行帧插值,但在含有强光源的情况下插值效果不理想,因此提出了针对强光源视频的帧插值方法。提取图像的光强特征,将平衡过光强的图像用于帧插值,将帧插值的结果恢复到初始光强。结果表明:在不含有高光强区域的视频中,插值速度及精度与目前最优方法相似;在含有高光强区域的视频中,帧插值精度高于其他方法,验证了该方法的有效性。Video frame interpolation is to improve video frame rate by synthesizing non-existent frames between original frames.The common frame interpolation method is based on depth perception,but the interpolation effect is not ideal in the case of high light source.A frame interpolation method for video with high light intensity is proposed.The light intensity features of the image is extracted,and then the image with balanced overlight intensity for frame interpolation,and the result of frame interpolation to initial light intensity is restored.The results show that the speed and accurary of the frame difference in the video without high light intensity region are similar to that of the current optimal methed,and the accuracy of the frame dfference in the video with high light intensity region is higer than other methods,which verifies the effectiveness of the proposed method.

关 键 词：视频帧插值 强光源 深度感知 光流 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]