Hi!

I don't know about saving background for background subtraction but maybe another solution to the problem would be to use a Kinect with pix_freenect (1) and create an alphamask out of the depthmap using pix_threshold_depth (2). you can mutliply this with the (aligned) rgb image from pix_freenect and you have a quite good extraction of the objects in the defined range. you can do these operations also on gpu with shaders which is more cpu efficient.

matthias

(1) https://github.com/kronihias/pix_freenect (2) https://github.com/kronihias/pix_threshold_depth

Am 05.04.12 12:21, schrieb Iain Mott:

Hello,

I'm working on an installation that will have stable lighting - and I'll be using pix_background in combination with pix_video to capture and isolate the image of individual participants and mix the participant's image (now with a black background) with other video sources.

My question is, can the calibration of pix_background (achieved by sending it a "reset" message while capturing the plain background without participant) be saved to file and later reloaded - or will I need to perform the "reset" calibration each time Pd is started?

Can this be done simply or will I need to delve into the C code? Has anyone tried this?

Cheers,

Iain

---

Pd-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list