Traditional broadcast productions typically utilise multiple cameras, yet broadcast only a single view of the scene they capture. Camera selection – decisions about when and how to cut – is performed by highly skilled human professionals, in cooperation with directors, camera operators and other members of the production team. The result is a single output stream where every viewer gets to see the same content. In contrast, a format-agnostic approach uses a scene capture model, and as automation comes into play, the result is less restricted. One implication is that users do not have to get the same view of the scene, and viewpoint selection could also be influenced dynamically by the users.
Automation of that view selection process is a fundamental research challenge for progress towards interactive TV. Its focus is to support a high degree of interactive content access and personalisation. Such software, usually referred to as a Virtual Director, has the potential to lead to more immersive user experiences, to unlock new content consumption paradigms, to increase the level of freedom for viewers in influencing what they get to see and to save production cost by generating content for a number of playout channels in parallel.
This chapter discusses how to further process content analysis results and manual annotations as described in detail in Chapter 5, for the sake of production ...