基于线性回归分析的视频质量评估方法

A Computationally Efficient Video Quality Assessment Method Based on Linear Regression Analysis

文章提出了一种利用线性回归分析方法的无参考视频序列质量评估算法。该算法主要利用帧间编码帧的比特数和该帧与其参考帧的差异两个参数进行线性回归分析来评估视频质量。该方法不需要原始参考视频,算法简单。通过对标准视频序列的仿真实验,该算法可有效评价不同视频的编码质量,使用该质量评估方法测得的失真视频客观质量评分与其主观质量评分有很好的一致性。

Sections 1 and 2 of the full paper explain the assessment method mentioned in the title, which we be- lieve is more computationally efficient than existing ones. Section 1 briefs the state of the art. Section 2 is entitled ＂No-Reference Video Sequence Quality Assessment Model＂ ; it needs to be divided into four subsections. Its core consists of： （1） we employ two parameters： the bit rate for each inter-coded frame and the difference between the inter-coded frame and its reference frame; （2） we carry out the linear regression analysis of the two parameters to assess the quality of a video sequence. Section 3 did experiments on five standard CIF video sequences to verify the effectiveness of our assessment method ; the experimental results, given in Figs. 3 and 4 and Table 1, and their a- nalysis show preliminarily that： （ 1 ） our assessment method is indeed more computationally efficient than existing ones and does not require the original video sequence for reference; （2） the assessment results obtained with our assessment method, which falls into the category of objective quality assessment, are very close to those obtained with the subjective assessment method.