No-reference video quality metric for HDTV based on H.264/AVC bitstream features

Abstract

No-reference video quality metrics are becoming ever more popular, as they are more useful in real-life applications compared to fullreference metrics. Many proposed metrics extract features related to human perception from the individual video frames. Hence the video sequences have to be decoded first, before the metrics can be applied. In order to avoid the decoding just for quality estimation, we therefore present in this contribution a no-reference metric for HDTV that uses features directly extracted from the H.264/AVC bitstream. We combine these features with the results from subjective tests using a data analysis approach with partial least squares regression to gain a prediction model for the visual quality. For verification, we performed a cross validation. Our results show that the proposed no-reference metric outperforms other metrics and delivers a correlation between the quality prediction and the actual quality of 0.93.Index Terms-H.264/AVC, HDTV, 1080p25, subjective testing, visual quality, video quality metric, no-reference metric.

Length-independent refinement of video quality metrics based on multiway data analysis

“…The dimensionality of the three-way feature array was reduced by temporal pooling as discussed in [1]. On the one hand the results confirms the results from [3] -the PLS1 metric produces similar correlation values for the dataset used here. On the other hand this again shows that the addition of the temporal dimension leads to better results.…”

“…In comparison to the no-reference metric by Brandão and Queluz [11] based on the same dataset, our metric shows better correlation values for quality estimation on the used dataset. In addition, we performed quality estimation on the same dataset with a PLS1-based metric similar to the one described in [3]. The dimensionality of the three-way feature array was reduced by temporal pooling as discussed in [1].…”

“…There have already been some contributions using a data analysis approach to build a no-reference video quality metric. Our most promising metrics are based on the extraction of H.264/AVC bitstream features from the video data which are used to train a regression model [1,2,3]. Recently we have shown that the inclusion of the temporal dimension by using multiway data analysis further improves the prediction [1].…”

“…Rossholm and Lövström [9] used bitstream features to estimate some other quality metrics additional to the PSNR. Keimel et al [1,2,3] refined the usage of bitstream features in order to directly estimate the visual video quality.…”

No-reference image and video quality assessment: a classification and review of recent approaches

“…A further improvement is found in [155] where a larger features set is used and the prediction of subjective MOS is also performed. A set of 48 bitstream parameters related to slice coding type, coding modes, various statistics of motion vectors, and QP value was used in [156] to predict the quality of high-definition television (HDTV) video encoded by H.264/AVC. PLSR was used as tool for regression between the feature set and subjective assessment.…”