基于暗通道及双流卷积网络的监控视频质量评估算法

Quality Evaluation of Surveillance Video Based on Dark Channel and Two-Stream Convolutional Network

受光照、气象、污损、视频压缩以及网络传输的影响,视频监控的质量可能会大幅降低,给后继的信息获取造成困难。为实现对监控视频质量的精确评估,本文提出了基于暗通道及双流卷积网络的监控视频质量评估算法,首先计算监控视频关键帧的暗通道图像,然后与原始图像合并,并利用双流卷积网络进行监控视频质量评估。该算法可准确评估监控视频中出现的模糊、块效应、噪声等质量缺陷。

Due to the influence of light,weather,dirt,video compression and network transmission,the quality of video monitoring may be greatly reduced,making it difficult to obtain subsequent information.In order to realize the accurate evaluation of monitoring video quality,this paper proposes a monitoring video quality evaluation algorithm based on dark channel and two-stream convolutional network.Firstly,the dark channel image of the key frame of monitoring video is calculated,then combined with the original image,and the monitoring video quality is evaluated by two-stream convolutional network.The algorithm can accurately evaluate the quality defects such as blur,blocking effect and noise in surveillance video.