Smartphones and tablets promote a personalized consumption of video content , making the democratic production of targeted content an exciting challenge for content providers. This research builds on computer vision tools to face this challenge in controlled scenarios, as most notably encountered for sport events.
As depicted in the figure below, the automatic production is in charge of two main decisions, namely camera selection and image cropping parameters.
In short, the production decisions in a static-camera infrastructure are derived from player (and ball) detection tools , so as to optimize a trade-off between three factors:
Completeness counts the number of salient objects in the scene, and thereby measures the integrity of the camera/viewpoint selection;
Fineness refers to the amount of details provided about the rendered scene. It is measured in terms of pixel resolution of the salient objects, and thereby favors close views selection. Increasing the fineness of a video does not only improve the viewing experience, but is also essential in guiding the emotional involvement of viewers through the use of close-up shots.
Smoothness refers to the graceful displacement of the camera viewpoint, and to the continuous story-telling resulting from the camera switching. Preserving smoothness is important to avoid distracting the viewer from the story with abrupt changes of viewpoint.
Based on the related scientific contributions ,, a fully automatic sport production solution has been developed within a European research project (www.apidis.org). The solution is now commercialized by Keemotion (www.keemotion.com) to cover team-sport events in a cost-effective manner, making the creation of video reports affordable, even in case of small or medium-size audience.
The following figures present two image samples resulting from the patented technology developed in those projects. In the first image, the technology is used to decide how to crop images in a panoramic view of the basket-ball field. In the second image, the production decisions include both camera selection and definition of image cropping parameters, as a function of the scene at hand.
Fan Chen, Damien Delannay, Christophe De Vleeschouwer, An autonomous framework to produce and distribute personalized team-sport video summaries: a basket-ball case study, IEEE Transactions on Multimedia, Vol. 13, no. 6, p. 1381-1394, December 2011. PDF
Chen, Fan ; De Vleeschouwer, Christophe. Personalized production of basketball videos from multi-sensored data under limited display resolution. In: Computer Vision and Image Understanding, Vol. 114, no. 6, p. 667-680, 2010. PDF
Christophe De Vleeschouwer. Autonomous infrastructures for networked interactive and personalized media experience. IEEE Communications Society, Multimedia Communications Technical Committee, E-letter, Vol. 6, no. 1, p. 24-27, January 2011. PDF
Pascaline Parisot, Berk Sevilmis, Christophe De Vleeschouwer, Training with corrupted labels to reinforce a probably correct teamsport player detector. Advanced Concepts for Intelligent Vision Systems, Poznan (Poland), October 2013. PDF