A distortion-agnostic video quality metric based on multi-scale spatio-temporal structural information
作者：
Highlights：
• Structural degradations due to video distortions considerably affect the perception of visual quality.
• LBP-TOP histogram is an effective way to measure the structural degradations in videos.
• Weighting the LBP using local magnitude based LBP significantly improves the performance without increasing histogram length.
• Multi-resolution analysis incorporates coarse-to-fine structures in quality assessment and enhances prediction accuracy.
• The proposed metric is computationally inexpensive and exhibits high correlation with the subjective ratings.
摘要
•Structural degradations due to video distortions considerably affect the perception of visual quality.•LBP-TOP histogram is an effective way to measure the structural degradations in videos.•Weighting the LBP using local magnitude based LBP significantly improves the performance without increasing histogram length.•Multi-resolution analysis incorporates coarse-to-fine structures in quality assessment and enhances prediction accuracy.•The proposed metric is computationally inexpensive and exhibits high correlation with the subjective ratings.
论文关键词：Video quality assessment (VQA),No-reference (NR),Local binary patterns (LBP),Three orthogonal planes (TOP),Artificial neural network (ANN),Video quality dataset
论文评审过程：Received 21 July 2018, Revised 23 February 2019, Accepted 24 February 2019, Available online 6 March 2019, Version of Record 15 March 2019.
论文官网地址：https://doi.org/10.1016/j.image.2019.02.010
