Adaptive streaming of multi-view video over P2P networks S. Sedef Savas a,n , C. G¨ oktu˘ gG¨ urler a , A. Murat Tekalp a , Erhan Ekmekcioglu b , Stewart Worrall b , Ahmet Kondoz b a College of Engineering, Koc - University, 34450 Sarıyer, Istanbul, Turkey b I-Lab Multimedia Communications Research, University of Surrey, Guildford GU2 7XH, UK article info Available online 16 February 2012 Keywords: P2P overlay networks Multi-view video Adaptive streaming Scalable video chunks 3D video perception abstract In this paper, we propose a novel solution for the adaptive streaming of 3D representa- tions in the form of multi-view video by utilizing P2P overlay networks to assist the media delivery and minimize the bandwidth requirement at the server side. Adaptation to diverse network conditions is performed regarding the features of human perception to maximize the perceived 3D. We have performed subjective tests to characterize these features and determined the best adaptation method to achieve the highest possible perceived quality. Moreover, we provide a novel method for mapping from scalable video elementary stream to torrent-like data chunks for adaptive video streaming and provide an optimized windowing mechanism that ensures timely delivery of the content over yanlıs - gibi. The paper also describes techniques generating scalable video chunks and methods for determining system parameters such as chunk size and window length. & 2012 Elsevier B.V. All rights reserved. 1. Introduction Stereoscopic videos have already made an impact on the multimedia industry. The market share of stereoscopic 3D movies in Hollywood has increased significantly over the last few years. More recently, stereoscopic TV broad- casts over DVB have begun in the UK in April 2010 using a frame-compatible format, which combines right and left views of a stereoscopic video pair in a single HD video frame. 3D-compatible TV sets that can display stereo- scopic videos using various technologies, such as using polarization or time-shutter glasses, are available in the market. Meanwhile, standards for transmitting 3D media signals over peripherals, such as HDMI 1.4a, have been defined. We are now progressing to the next phase in 3D media services, which will be based on multi-view-video (MVV) formats. MVV enables viewing a scene in 3D from multi- ple angles (within a viewing cone), which may make it possible to see behind an object by tilting the head, whereas in stereoscopic video, the viewer can only see what has been captured from a single viewpoint. More- over, MVV displays are auto-stereoscopic which do not require wearing special 3D glasses. The major difficulty with MVV is that the amount of data, and hence the bit- rate required to transmit it, depends on the number of views; making it difficult to transmit over fixed bit-rate channels, such as the DVB. Transmission of MVV over IP (Internet Protocol) is perhaps the most flexible solution for 3D media delivery, which can provide different trans- mission rates to different users according to their avail- able connection rate and display technology. In addition, it is also possible to receive feedback from the user over an IP channel, enabling effective rate adaptation in addi- tion to personalized services such as user centric adver- tisements or interactive TV. The MVV over IP can be offered as a stand-alone service or as a supplement to broadcast of stereoscopic video over DVB. It is well known that video over IP has some limita- tions. First, IP operates in the best effort sense; hence, it is possible to experience packet losses and varying amount Contents lists available at SciVerse ScienceDirect journal homepage: www.elsevier.com/locate/image Signal Processing: Image Communication 0923-5965/$ - see front matter & 2012 Elsevier B.V. All rights reserved. doi:10.1016/j.image.2012.02.013 n Corresponding author. E-mail addresses: Ssavas@ku.edu.tr (S. Sedef Savas), Cgurler@ku.edu.tr (C. G ¨ oktu˘ gG¨ urler), mtekalp@ku.edu.tr (A. Murat Tekalp), E.Ekmekcioglu@surrey.ac.uk (E. Ekmekcioglu), S.Worrall@surrey.ac.uk (S. Worrall), A.Kondoz@surrey.ac.uk (A. Kondoz). Signal Processing: Image Communication 27 (2012) 522–531