Hacks from the kitchen

Greetings!

GREAT VIDEO! To tell you the truth, I was awestruck to see it the first time. Almost fell off my chair, literally. Just have a few queries here friend. Compared to the original video, how much size in bytes did you save in recording it? I almost thought you already have a running vector-video capture engine or something.... But for that to happen you would still have to take raw input from the device (or pc) DAC and still have the non-vectorized input and convert is still. Had the same idea also.... That would be totally, TOTALLY cool!!! I guess you know what implications/applications of that concept mean.

THANKS!

onezie

Have you tried grouping the video into quasi-objects? I don't mean saying "this is a chair, I'll isolate it", but saying "this region of pixels seems to update independently from the rest of the video, so when I vectorize it I'll make it its own group in the output" (since most vector formats support grouping). That way, if you want to convert a bitmap video into swf, for example, you can use those groups to make tweens where the contents of the object doesn't change but its location does. That would save a decent amount of space in normal video, and an amazing amount of space in cartoons/anime, where the video is created that way in the first place.

This works strictly on stills - so there's no use of movement at all.
I'm - unrelated to this - working on some 3d perspective stuff and what you talk about might be useful there.

There's the very cool VideoTrace project which, sadly, is not yet available for download/use: http://www.acvt.com.au/research/videotrace/