That's very interesting. It's too bad that passing that negated image into an OpenCL kernel clips the pixels to non-negative values. Is that necessary? Seems like a feature request.
Actually, no CL was harmed in the making of this composition (I was in Leopard for the time being, sadly, so no CL love here) -- that was just a basic Core Image filter.
Of course, I'm not interested in saving negatives pixels to a file. They're just variables in a computation.
As long as you're able to generate them somehow, CoreImage is fine with using/passing along negative numbers. I'm pretty sure (though haven't tested) that sending a negative-value-containing image to the CL Kernel you're speaking of would do exactly what you expect -- move pixels up or left, instead of just down and right.
The trick, ultimately, is in generating negatives -- most image sources will be positive, so you'll likely have some scale/bias step along the way to get data ranged as you want.
-- [ christopher wright ] [email protected] http://kineme.net/
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ Do not post admin requests to the list. They will be ignored. Quartzcomposer-dev mailing list ([email protected]) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/quartzcomposer-dev/archive%40mail-archive.com This email sent to [email protected]
