On Feb 18, 2011, at 11:04 PM, Mathieu Bouchard wrote:
On Sat, 19 Feb 2011, Matteo Sisti Sette wrote:
Le 18/02/2011 19:23, Hans-Christoph Steiner a écrit :
I don't think there is a driver that'll just give you the video and the depth. Its like a video camera with two images per frame: the normal video frame and the depth frame.
Wouldn't it be possible to have the depth value in the alpha channel and so obtain a single rgba image?
Video4Linux1's VIDEO_PALETTE does not mention any alpha format.
Video4Linux2's V4L2_PIX_FMT does not mention any alpha format either.
In Video4Linux2's V4L2_FBUF_FLAG there are three things called alpha. They're related to overlay. That's a thing for capture cards that can write directly to the graphics card.
There's nothing really for 4 channels... well, maybe there is. There is RGGB, BGGR, GBRG, you know, the Bayer stuff that has double the resolution for the G channel, that is used in "raw" cameras. A Video4Linux driver could use that fourcc.
But I don't know whether there are any restrictions to having kernel drivers create their own fourcc as long as an app accepts it.
However, current linux support for the Kinect doesn't use any kernel drivers, right ? Both existing libs use libusb, which is userspace.
I believe that there is both: the libusb/libfreenect approach, which will work on GNU/Linux, Mac OS X, and perhaps also Windows, and the Linux kernel driver approach.
.hc
----------------------------------------------------------------------------
"A cellphone to me is just an opportunity to be irritated wherever you are." - Linus Torvalds