Fine Granularity Adaptive Multi-Receiver Video Streaming Viktor S. Wold Eide, Frank Eliassen Jørgen Andreas Michaelsen, and Frank Jensen viktore@simula.no frank,jorgenam,fnjensen@ifi.uio.no Simula Research Laboratory and University of Oslo, Norway MMCN’07 San Jose, CA, USA, 2nd of February 2007 [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 1 / 15 Introduction Outline Background Adaptation for Fine Granularity Multi-Receiver Video Streaming Experimental results Conclusion and further work [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 2 / 15 Background Fine Granularity Multi-Receiver Video Streaming A video coding and streaming scheme we proposed in MMCN’05 article: multi-receiver streaming each receiver may customize video quality independently receiver-driven scalable: sender side network receiver side [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 3 / 15 Background Fine Granularity Multi-Receiver Video Streaming Using techniques from MPEG, but designed for scalability A GOP is divided into fragments that each contribute to a quality dimension Each dimension is independent, and each dimension is layered temporal luminance quality chrominance quality regional Publish-subscribe based video streaming receivers customize the quality by subscribing to a certain number of layers in each dimension [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 4 / 15 Demonstration [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 5 / 15 Fine Granularity Adaptive Multi-Receiver Video Streaming [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 6 / 15 Challenge The video quality dimensions are independent a combinatorial explosion in potential ways to adapt maintain expressiveness in underlying scheme consequently, an automatic solution is necessary Receiver-driven Automatic Adaptation Coding and Streaming Scheme for Fine Granularity MultiReceiver Video Streaming [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 7 / 15 Adaptation The adaptation decision process User preferences Adaptation Decision Resource availability Video quality subscription Coding and Streaming Scheme for Fine Granularity MultiReceiver Video Streaming [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 8 / 15 Resource Availability The resources considered in this work are Network bandwidth CPU utilization [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 9 / 15 Utility Functions User preferences by means of utility functions U (t , y , c ) = N X Ui (t , y , c ) (1) i =0 Ui (t , y , c ) = Yi0 (t , y ) + Ci0 (t , c ) (2) Yi0 (t , y ) = Yi (y ) + sgn (Yi (y )) Ti (t ) (3) Ci (t , c ) = Ci (c ) + sgn (Ci (c )) Ti (t ) (4) 0 Wt Wit Kt (t ) (5) Yi (y ) = y Wy Wi Ky (y ) (6) Ci (c ) = Wc Wic Kc (c ) (7) Ti (t ) = [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 10 / 15 Experiments Adaptation scheme evaluated for different user preferences: Action — frame rate most important Detail — frame quality most important Parallel processing — some superblocks, some part of video signal [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 11 / 15 Experiments: User prefer details, spatial quality Adaptation to available bandwidth full +3 layer +1 50 base TL QY QC utility null 0 utility (%) 75 +2 25 none 200 400 600 800 1000 1200 1400 BW (kbps) [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 12 / 15 Experiments: User prefer details, spatial quality Utility as a function of resource availability [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 13 / 15 Experiments: User prefer details, spatial quality Resource utilization, normalized metric [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 14 / 15 Summary An Adaptation Scheme for Fine Granularity Multi-Receiver Video Streaming Fine grained adaptation is feasible given radically different user preferences Both resource utilization and user utility can be kept at a high level Adapts to variations in available bandwidth and CPU resources, roughly over two orders of magnitude Further work Quality of adaptation — addressing when and how to adapt Adjust quality based on visual focus [simula . research laboratory ] Viktor S. Wold Eide (Simula, Ifi UiO) Fine Granularity Adaptive Video Streaming 2nd of February 2007 15 / 15