Forums > Jitter

kinect depth map smoothing on gpu

Feb 20 2013 | 2:20 am

i know there is a bunch of posts and patches on this but most are on the cpu and aimed at transforming 3d-stuff in the end.
in my case i want to use the depth map as an alpha matte, keying out all people close to camera – so i need a matte as smooth and nice as possible.
below i attached what i gathered so far: depth selection followed by dilate, erode and blur. it works ok but i guess it is missing some temporal smoothing (temporal median or something? i tried slide but that did not help) and edge smoothing (like you can in the mask settings of postproduction programs).
can anyone help me? or maybe point me the way, what sort of operation i would need to look out for?
thanks and all the best!

-- Pasted Max Patch, click to expand. --

Feb 22 2013 | 9:09 pm

i am still looking for a way of temporal smoothing of my kinect depth map… can anyone help me?
i inserted a slide shader and it does help a little but i think it’s just the wrong approach…
what would be the keywords/methods i have to look out for?
or is there any gpu-version of calculating the temporal median (i read that in the forum somewhere but actually all pasted patches were gone…)?
attached an updated version of my current processing chain.

-- Pasted Max Patch, click to expand. --

Feb 22 2013 | 9:58 pm

Hi Karl,

I, too, am looking for a way to "smooth" or simplify/optimize/minimize video and/or vector points in Max/Jitter, so I would be interested to know what you learn about this process. Something like the "Smooth selection" function in Photoshop!

I’ll let you know if I stumble upon something like this…

Viewing 3 posts - 1 through 3 (of 3 total)

Forums > Jitter