Television video-effects are implemented by a video production switcher (VPS). A VPS can switch from n inputs to one output that is typically the broadcast output. A VPS can apply effects on the input(s) as the data is routed to the output. Commercial systems use custom-designed hardware to meet the processing requirements of broadcast quality video. Some VPS systems can layer effects on top of other effects (e.g., fading from one stream to another with titling). However, the number of layers is bounded, and most VPS systems do not allow new effects to be added.
The Parallel, Software-only Video-effects Processing System (PSVP), which runs on a Network of Workstations, was developed by our research group to provide a low-cost, extensible video-effects system [10]. PSVP produces an effect by decomposing it into a collection of processes that read RTP packets from a multicast session, compute an effect, and write RTP packets back into the multicast session.
Dc invokes and controls an effect through the service infrastructure. Effects services (see figure 2) input streams from the studio session, render an effect, and output a new ``effects stream." Effects streams are ordinary sources as far as dc is concerned with controls for the effect. Thus, effects can be cascaded by piping the output of an effect to the input of another effect service. A titling effect is shown in the Berkeley MIG Seminar webcast shown in figure 1. Adding a new effect to the webcast requires only that a new service be installed in the infrastructure.