1. New requirement of subjective video quality assessment methodologies for 3DTV Wei Chen(*) 1,2, Jérôme Fournier 1, Marcus Barkowsky 2, Patrick Le Callet 2 1 Orange labs, France télécom R&D 2 IRCCyN IVC Team, Ecole Polytechnique de Nantes Orange labs R&D - IRCCyN IVC Team1VPQM 2010

2. Motivation A high quality 3DTV broadcast service is becoming more and more realistic. VPQM 2010Orange labs R&D - IRCCyN IVC Team2 End-user Visual quality Excellent Bad Good Fair poor Naturalness Presence 3D perception visual experience Visual Comfort eye strain etc. Headache etc. Network Mobile (WiMax, 3GLTE) xDSL, FTTx, … Codec Proprietary Standardised New generations MPEG-C, MVC, LDV Terminal Anaglyph system Active Shutter Glass Glasses system Passive (Ploarised) Glasses system Autostereoscopic system Production AV acquisition Post-production 2D/3D conversion Multiview interpolation View synthesis from 3D representations Mixed reality (augmented reality) 3D/3D conversion … 2D+z Multi-view Others Coming generations DES

3. Subjective video quality assessment – Evaluation the human perception – Optimize 3DTV systems – Ease the specification process for end to end application – A solid reference for objective quality metrics. – Comparison of the results from different laboratories However, 2D subjective video quality assessment can not fulfill this requirements. VPQM 2010Orange labs R&D - IRCCyN IVC Team3

4. Content 1.Novel condition and requirements(selected) of 3DTV regarding subjective quality assessment. 2.Depth rendering of 3DTV – A new notion: depth rendering ability – Comparison of different 3D displays 3.Conclusion VPQM 2010Orange labs R&D - IRCCyN IVC Team4

5. General Viewing Conditions Luminance and contrast ratio Screen luminance ≠ Perceived luminance crosstalk affects contrast ratio VPQM 2010Orange labs R&D - IRCCyN IVC Team5 -80% -60%

6. General Viewing Conditions Monitor resolution Recommendation of minimum values for spatial per view resolution VPQM 2010Orange labs R&D - IRCCyN IVC Team6 1080i/p 960x1080 L R Line Interlaced 960x1080 1920x540 LR Column Interlaced 1920x540 640x 360 LR Multi-views 640x 360 L R Full resolution 1920x1080

7. General Viewing Conditions Background and room illumination Minimum distance between display and background necessary Technology of room illumination critical VPQM 2010Orange labs R&D - IRCCyN IVC Team7 ? Wall TV Neon illumination

8. General Viewing Conditions Viewing distance Adding depth perception factor into Preferred Viewing Distance – Depth of Focus depends on viewing distance – Depth rendering also relates viewing distance Preferred Viewing Distance sometimes fixed by display manufacturer – for example, autostereoscopic displays Philips 423D6W 42 inch (3 meters=5x height) Alioscopy 40 inch (more than 4 meters = 7x height) VPQM 2010Orange labs R&D - IRCCyN IVC Team8

9. General Viewing Conditions Depth rendering Human perceptual factor – Threshold for Depth Of Focus and binocular disparity. 0.2 diopter or 0.3 diopter DOF 60 arcmin binocular disparity VPQM 2010Orange labs R&D - IRCCyN IVC Team9 +0.2 diopter-0.2 diopter Comfortable viewing zone I can not fuse the image and I feel uncomfortable.

10. Test method Subjectively measured quality indicator – Additional indicator besides image quality, e.g. Naturalness, Presence, Visual Experience Visual discomfort – Questionnaires for visual discomfort – objective measurement of visual fatigue VPQM 2010Orange labs R&D - IRCCyN IVC Team10

11. Depth rendering of 3D Display VPQM 2010Orange labs R&D - IRCCyN IVC Team11 Stereoscopic voxel 0-2-3-41234 Depth plane 65mm viewing distance pixel

12. Definition of Depth Rendering Ability the number of depth planes that can be represented within a comfortable viewing zone of display. Parameters affect the depth rendering ability: – Size of comfortable viewing zone – Maximum uncrossed disparity in pixels – Angular depth plane interval Parameters affect immerse experience – Field of view(Horizontal visual angel) VPQM 2010Orange labs R&D - IRCCyN IVC Team12

13. VPQM 2010Orange labs R&D - IRCCyN IVC Team13 e d viewing distance 0-2-3-41234 Comfortable viewing zone Maximum uncrossed disparity in pixels Depth rendering ability in pixels Angular depth interval Screen Filed of view

14. Comparison of the depth rendering abilities for different displays VPQM 2010Orange labs R&D - IRCCyN IVC Team14

15. Comparison of the depth rendering abilities for different displays VPQM 2010Orange labs R&D - IRCCyN IVC Team15

16. Comparison of the depth rendering abilities for different displays VPQM 2010Orange labs R&D - IRCCyN IVC Team16 Visual threshold 0.8 arcmin Visual threshold 0.8 arcmin Visual threshold 30 degree Visual threshold 30 degree

17. Discussion Best system discussed: – 2 projector Full resolution solution 100 inches. Small size display needs larger comfortable viewing zone – longer viewing distances may be useful Autostereoscopic display may show depth quantization effect – longer viewing distances or higher resolution panel may be advised. 2K cinema screen have poor depth rendering ability – higher resolution(now 4K, future 8K) may be strongly suggested. Depth rendering abilities vary largely(*) – Content disparity should be adjusted to the display used. VPQM 2010Orange labs R&D - IRCCyN IVC Team17

18. Cinema Content to Home Display VPQM 2010Orange labs R&D - IRCCyN IVC Team18 Cinema screen 8 meter height infinite 8 pixels14 pixels Directly use 4.9 m 43 pixels infinite 8 pixels 43 pixels Poor depth perception 1.78 m 16 m 2 projectors screen 1.35 meter height

19. Conclusion We discussed about various new requirements for subjective video quality assessment of 3DTV. – The requirement of new subjective methodology is urgent! We proposed an novel notion depth rendering ability combing physical parameters and perceptual constrains of 3D display. – identify the 3D display depth rendering performance – select the appropriate content for certain display VPQM 2010Orange labs R&D - IRCCyN IVC Team19

20. Thank you! Question? VPQM 2010Orange labs R&D - IRCCyN IVC Team20