Rotoscoping or matting out MUCH FASTER !!!

preety cool

It does look cool but I’m not sure they get a great edge from it. Fine detail retention is the primary goal of chromakeying.

Also does the camera need to be mostly stationary? It reminds me of the unwrapping startegy that allow you to paint onto infered geometry from a video clip.

well yeah, but this paper is from 2005. its been allready 7 years, i guess there is still room for improvements…

we need this in Blender

The result looks nice.
But all those weird extrusions are confusing :confused:

Another way to visualise elements is a wonderful thing (cube simulating time that is), but I have looked at depth layering I really doubt that this technique would yield good edges. However it would be a wonderful tool for garbage matteing, that is knocking out a rough mask to eliminate all the bad blue screen rubbish you don’t need without the trouble of hand roto.

Oh endi, endi…

Trippy, I imagine that’s what it would look like to perceive the 4th dimension.

Yeah, I was looking at the edges produced (the video was low res) but what I noticed wasn’t that great. I don’t know if we need this in Blender, as with the new masking able to track I think we could make very quick and easy garbage mattes using those techniques instead.

If I understand this correctly, it depends on object motion, implying to me that somewhat faster moving objects may provide better results. It also seems likely that first camera tracking would also help improve precision, especially in the elephant clip. The elephant clip is a good test because objects that move slowly, and are large and of a fairly uniform color, may be difficult to differentiate the center mass from stationary background.