I noticed something rather odd, which may explain my confusion with
DOD/ROI and sampler space.
My system Mac OS X 10.5.1, Intel MBP 2.33 with QT 7.3, I have the
following behaviour within QC.
Step 1:
Create a Core Image Kernel with the following code:
kernel vec4 v001AlphaBlend(sampler image1, sampler image2, sampler
mask)
{
// first we find out which image extent is the largest - we must do
this since CIKL does not
// normalize sampler coordinates for us like GLSL does.
// We need to know which sampler contained that largest size..
// mix based on masks alpha value
return mix(img1, img2, msk.a);
//return img1;
}
Step 2:
Pass three movies into the input, preferably those with different
sizes, for me, my mask movie, which contains an alpha channel, is
smaller than image input 1 and 2, and connect to a billboard.
Step 3:
Now, select the smaller movies playback patch. Enable/disable color
correction. With my system, Disabling Color Correction seems to
supply different coordinates to my CIKernel. The output is no longer
within the same place when color correction is enabled:
No CC:
<NoCC.jpg>
With CC:
<WithCC.jpg>
Can anyone verify this? Is my Kernel code incorrect? ive been going
crazy and then noticed I had disable CC checked on my mask movie,
and then things started to make sense...
