基于时空预测模型的视频码率控制算法  

Video Rate Control Algorithm Based on Spatiotemporal Prediction Model

作　　者：李汶隆 刘念林 柳春青 LI Wenlong;LIU Nianlin;LIU Chunqing(Sichuan Kerui Software Co.,Ltd.,Chengdu 610041,China;Sichuan Hushan Electric Appliance Co.,Ltd.,Chengdu 610041,China)

机构地区：[1]四川科瑞软件有限责任公司,四川成都610041 [2]四川湖山电器股份有限公司,四川成都610041

出　　处：《电视技术》2021年第6期115-118,共4页Video Engineering

摘　　要：针对传统视频图像码率控制算法不能很好地支持视频内容变化或者场景切换应用问题,提出一种基于时空预测模型的视频码率控制算法。该算法根据视频图像压缩时基于空间关联与时间关联的预测属性,引入人体视觉感知模型,通过对时间通道中的视频内容进行理解,采用视频图像帧间绝对误差均值(Mean Absolute Deviation,MAD)的方差来表征视频图像在时间通道中的运动趋势,评估当前视频图像帧内容复杂度并预测出MAD值;在此基础上,通过时空关联因子对MAD的预测值进行矫正,使得码率控制算法中关键参数MAD值能更好地贴合实际情况。实验证明,采用时空维度分析的码率控制算法更科学、更精准。Traditional video rate control algorithms can’t support video content change or scene switching well.This paper proposes a video rate control algorithm based on spatiotemporal prediction model.According to the prediction attribute based on spatial correlation and temporal correlation in video image compression,the human visual perception model is introduced.By understanding the video content in the time channel,the variance of Mean Absolute Deviation(MAD)between video frames is used to characterize the motion trend of video image in the time channel,the content complexity of the current video frame is evaluated and the MAD value is predicted;on this basis,the predicted value of MAD is corrected by the spatio-temporal correlation factor,which makes the MAD value of the key parameter in the rate control algorithm better fit the actual situation.Experiments show that the rate control algorithm based on spatiotemporal dimension analysis is more scientific and accurate.

关 键 词：码率控制 阈值 方差 时空关联因子 

分 类 号：TN919.81[电子电信—通信与信息系统]