On 03/06/2011 04:05 PM, tim vets wrote:
Dear list, When reading with [pix_video] from a camera via Webcamstudio (vloopback) at /dev/video1 the colors are incorrect. It looks like red and blue are inverted or swapped somehow, orange becomes blue and vice versa. Reading directly from the camera (driver 0 (v4l2), /dev/video2) the colors are correct. The same loopback device /dev/video1, v4lsrc is read by a Proce55ing sketch with correct colors.
it's a known bug which i believe is fixed in current SVN.
until the next release you could either try matju's suggestion, or try sending a "colorspace yuv" to [pix_video].
if this works, chances are high that you have a lower processing overhead than using [pix_colormatrix].
I tried inserting [pix_yuv], [pix_rgba], but without success.
because the color is already wrong. [pix_yuv] and [pix_rgba] are designed to give you (nearly) the same result, only in a different colorspace.
the trick above (requesting "yuv" from [pix_video]) works (well, if it does), because the image acquisition code for RGB32 is a broken while the YUV acquisition code is not. (using [pix_yuv] after [pix_video] is already too late)
unfortunately, my suggestion might as well _not_ work in your specific case, since loopback devices will often only ofter their format in one single colorspace.
gfadmsr IOhannes