Thanks Max for this example, in fact it is quite close to my first trial, it works fine but if there is 2 persons in the image, one move, the 2 persons appears and I would like to see only the person who is moving. I'll dig deeper...
Benjamin
Max Neupert a écrit :
hi benjamin,
sebastian trippner did exactly this in i seminar i gave. here is the
description and patch to download: http://kunstundmedien.burg-halle.de/LEF/index.php?labor=erkenntnis&s=LEF he was using a very simple method: a snapshot (still image) from the
video feed was placed over the video image as soon as the video
showed no more movement. it's a very simple patch with not many onbjects.have fun with it.
Am 01.03.2007 um 06:41 schrieb >------------------------------:
Hi list,
I' trying to make an effect on a live camera feed with GEM for an installation in which only persons who move appear on a kind of video mirror, as chameleonTv of effecTv (http://effectv.sourceforge.net/chameleon.html) I tryed severals ways :
with pix_background and pix_blob, modulating the alpha value of the
rectangle on which the video is textured in function of the
quantity of movement but if one person moves, everybody appears.with pix_multiblob and combination of pix_image pix_rectangle and
pix_mask, the [pix_multiblob 2] seems to eat all the processor,
even if I [pix_resize 320 240] the image before the multiblob.accumulating many (25) pix_movement and using the result by
pix_masking it with the live feed dig a hole in the person that
moves a little. Maybe I did not the "average" of alpha channel of the 25 previous frame as I thought, due to the fact that pix_movement
works on the previous frame ? is there a better way with opengl instructions ? I tryed to look in
that direction but I didn't manage yet to find how to analyze and
affect the pixels of a live feed.... thanks for any advice benjaminfor the tech note : the video feed is from a PCI capture card (AlchemyTv) connected in s-video to a dv camera with [pix_video 720
576] on a G5 10.4 Pd ext7, for the first trial, I manage to mix this video feed with another DV camera connected in firewire, putting into a buffer an image
background, filling and playing another buffer with PAL images, and playing a non compressed video of the same size >>> output on a second screen 800x600, with all that, the motion capture was really fast, only
40% of the proc used, it was really impressive (and stable), great dev
work, thks-- ^ -<[O:O]>- ~ +--[¨]--< | | _/ \_
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/ listinfo/pd-list
PD-list@iem.at mailing list UNSUBSCRIBE and account-management -> http://lists.puredata.info/listinfo/pd-list