论文部分内容阅读
针对广泛用于测量视频信号质量的指标峰值信噪比(PSNR)因需源信号而受限的问题,提出一种无需参考源的PSNR估计方法。该方法从视频码流中提取离散余弦变换(DCT)系数的量化值进行统计,得到其概率密度函数,并以此依据Laplace模型估计PSNR。该方法针对最新的视频压缩标准H.264的帧结构、变换和量化结合方式做出了优化,使得不同的编码速率和帧类型的码流均可适应。使用不同码率和不同运动方式的H.264码流进行测试,比较使用新方法得到的PSNR与实际PSNR间的差异。结果显示CIF格式的中低码率情况下二者的均值在1 dB以内。
In view of the problem that PSNR, which is widely used to measure the quality of video signal, is limited due to the source signal, a PSNR estimation method without reference source is proposed. The method extracts the quantized values of discrete cosine transform (DCT) coefficients from the video code stream to obtain the probability density function, and then estimates the PSNR according to the Laplace model. This method is optimized for the frame structure, transform and quantization combination of the latest video compression standard H.264, so that the code streams of different coding rates and frame types can be adapted. The H.264 code stream with different code rates and different motion modes was used to test the difference between the PSNR obtained with the new method and the actual PSNR. The result shows that the mean of the two under the low bit rate of CIF format is less than 1 dB.