H.264/AVC网络视频编码失真评估的比特流层模型

为了实现对采用H.264/AVC标准编码的网络视频质量的实时监测,提出一种评估视频编码质量的比特流层模型.该模型无需完全解码,只通过简单解析视频的量化参数、编码比特率以及运动矢量等信息评估视频流质量.首先,通过主观评估实验分析确定量化参数与视频编码失真的基本关系模型,然后利用量化参数和编码比特率预测视频的空间复杂度,利用运动矢量信息预测视频的时间复杂度,并结合空域掩盖效应和时域掩盖效应建立起一个能够反映人视觉特性的视频质量评估模型.实验结果表明,与解析码流的无参考网络视频质量评估模型相比,利用该模型得到的客观质量分数与主观质量分数的皮尔森相关系数提高了0.0160,均方根误差下降了0.0797.

A Bitstream-Layer Model for Assessing Coding Distortion of H.264/AVC Networked Video

A bitstream-layer model for video coding distortion assessment is proposed to realize real-time quality monitoring for H.264/AVC networked video.The model realizes real-time quality assessment for networked video without resorting to full decoding,but by simply analyzing the payload information such as the quantization parameters,bit-rate and motion vectors.Firstly,the relationship between the coding distortion and quantization parameters for different video sequences is modeled.Then,the quantization parameters and bit-rate are employed to predict the spatial complexities,and the motion vectors are employed to predict the temporal complexities.After empirical data analysis,a mathematical model for video quality assessment is established taking account of the spatial and temporal masking effect of the human visual system.Experimental results show the advanced performance of the proposed model.A comparison with the NR QANV-PA model shows that the proposed model gets an increment about 0.0160 in the Pearson Correlation Coefficient between the objective scores and mean opinion scores and a decrement about 0.0797 in the root mean square error.