计算机应用研究2018，Vol.35Issue(2)：611-614,4.DOI:10.3969/j.issn.1001-3695.2018.02.062
基于卷积网络的帧率提升算法研究
CNN-based frame rate up-conversion algorithm
摘要
Abstract
Motion compensated-based frame rate up conversion (MC-FRUC) is a primary method for frame rate up conversion.This paper proposed a new self-learning-based frame rate up-conversion (FRUC) algorithm via CNN to decrease the block artifact,occlusion and holes problem in the interpolated frame.The role of CNN was to predict the intermediate frame by two adjacent frames.This article assumes that the high frame rate sequence was available in the training phase.It trained the network parameters using high frame rate and low frame rate video.Finally,it stored or transferred the data in the form of video plus network parameters.In fact,this made the video provider bear larger burden to achieve the convenience of the video receivers.This was the key to improve the user experience for the video site.In experiments using benchmark image sequences,the proposed algorithm improves the average peak signal-to-noise ratio of interpolated frames at least 0.6 dB when compared to conventional motion estimation algorithms.And the proposed method can effectively avoid the block effect,hole and occlusion problems thanks to this approach is global prediction-based method.关键词
卷积神经网络/帧率提升/自学习Key words
CNN/FRUC/self-learning-based分类
信息技术与安全科学引用本文复制引用
侯敬轩,赵耀,林春雨,刘美琴,白慧慧..基于卷积网络的帧率提升算法研究[J].计算机应用研究,2018,35(2):611-614,4.基金项目
国家自然科学基金资助项目(61402034,61210006,61501379) （61402034,61210006,61501379）
北京市自然科学基金资助项目(4154082) （4154082）
中央高校基本科研基金资助项目(2015JBM032) （2015JBM032）
国家科技攻关计划资助项目(2016YFB0800404) （2016YFB0800404）
