Mobile Video Quality Assessment Database Lark Kwon Choi with Anush K. Moorthy, Prof. Alan C. Bovik, and Prof. Gustavo de Veciana
Outline Introduction LIVE Mobile VQA Database Subjective Study Source videos and distortion simulation Test methodology
Evaluation of Subjective Opinion Discussion and Conclusion 2
Growth in Mobile Video Traffic More devices, higher bit rates contents 78 % Global Mobile Data Traffic Growth by 2016
Mobile Video Traffic 70.5 %
How? Need for more video capacity, viewer’s QoE 3
Promising Direction “Perceptual optimization” of video networks Video Compression
Transmission
Visual Perception
Wireless Networks Feedback
Humans are final “receivers” of videos To understand human’s opinion and behavior on visual quality,
HD Mobile VQA Database & Subjective Analysis 4
Previous Subjective Studies Subjective studies
Focus
[K. Seshadrinathan et al., 2010] [A.K. Moorthy et al., 2010] [VQEG VQA Phase I and II, 2000, 2003] [VQEG Multimedia Phase I, 2008]
• Large displays • Distortion: Compression, IP/wireless loss
[S.R. Gulliver et al., 2007] [Q. Huynh‐Thu et al., 2006]
• Delayed and jitter
[A. Eichhorn et al., 2009] [H. Knoche et al., 2005] [S. Jumisko‐Pyykko et al., 2005, 2008] [M. Ries et al., 2007] [S. Winkler et al., 2003]
• Mobile devices
Limits • Results cannot be translated into small mobile devices
• • • • •
Small datasets Insufficient distortions Unknown source Small resolution Lack of publicity
To aid the development of perceptually optimized algorithms for wireless video transmission,
LIVE Mobile VQA Database 5
10 Reference, 200 distorted videos, and >50 subjects
Source Videos Digital Cinematographic Camera 12bit REDCODE RAW 2K (2048x1152), 30/60 fps
RED ONE 10 Actual Study 2 Training
Downsampled 720p (1280x720), Uncompressed YUV, 15 sec, Reference videos 7
Distortion Simulations H.264 Compression Wireless Packet Loss Frame freezes Rate adaptation Temporal dynamics 8
Distortions 4 Compression + 4 Wireless Packet Loss ‐ JM H.264 SVC / R1