[PD] Video morphing from live capture to pre-recorded video

[Charles Goyard]

Hi list,

at a general-culture class I gave about digital effects for stage
performances, 3 or 4 people asked about how to switch from live video to
pre-recorded video seamlessly. The context is in shadow theater, so
we're only dealing with contours/outlines/shapes.

The basic setup is that the shadow seen by the audience is in fact a
live video projection of the real shadow, captured with infrared
lighting and infrared camera. This permits "video scratching" of the
shadow, so it can detach from the actor/puppet, lag behind and then
catch up the movements of the actor/puppet. This works pretty well, but
at the moment I only have a vvvv implementation, not a pd one.

The idea is to add a means to transition into pre-recorded video, such
as, say, the shadow of bird flewing away.

I imagine such a setup with the following steps :
- capture the actor's shadow
- vectorise the outline of the shadow
- build paths to the target outline (the bird)
- stretch the actor's outline towards the bird's outline.

I guess there are some tools in opencv or some other lib for
vectorisation or outline detection.

Maybe it's easier to have some flood-fill algorithm to progressively
fill the gaps between the two outline.

I'd really like to build such a tool, because there's somewhat high
demand on that.

So the question is : what process makes senses, what libraries/objets
should I use to build such a tool ?

Any suggestions are welcome.

Cheers,

Charlot