研究目的
To introduce a new video quality model (VQM_VFD) that accounts for the perceptual impact of variable frame delays (VFD) in videos and to evaluate its performance on the LIVE mobile video quality assessment (VQA) database.
研究成果
The VQM_VFD model significantly contributes to the progress of VQA algorithms by accurately predicting human subjective judgments and outperforming existing top-performing IQA/VQA models. However, there remains significant room for improvement, particularly in handling temporal distortions and color distortions.
研究不足
The VQM_VFD model does not account for color distortions and is limited in its ability to handle temporal dynamics, as it does not take the order of events into account. Additionally, the model's performance is dependent on the accuracy of the variable frame delay (VFD) algorithm.
1：Experimental Design and Method Selection:
The study involved the development and testing of the VQM_VFD model, which uses perceptual features extracted from spatial-temporal blocks and a long edge detection filter to predict video quality. The model was evaluated against human subjective judgments of visual quality.
2：Sample Selection and Data Sources:
The LIVE Mobile VQA database was used, which contains a variety of video impairments typical of heavily loaded wireless networks, including dynamically varying distortions such as frame freeze and time varying compression rates, as well as static distortions such as compression and wireless packet loss.
3：List of Experimental Equipment and Materials:
The study utilized video sequences from the LIVE Mobile VQA database, which includes 720p 30 fps videos with a wide range of quality and examples of most common mobile video impairments.
4：Experimental Procedures and Operational Workflow:
The VQM_VFD model was tested on the LIVE Mobile VQA database to evaluate its efficacy at predicting human opinions of visual quality. A detailed correlation analysis and statistical hypothesis testing were conducted.
5：Data Analysis Methods:
The performance of VQM_VFD was analyzed using Spearman Rank Order Correlation Coefficient (SROCC), Pearson’s (Linear) Correlation Coefficient (LCC), and the root mean-squared-error (RMSE).
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
