I wrote a couple plug-ins for QC 3.0 (detailed on my QC blog: http://www.samkass.com/blog)
. They're not very efficient, and I was wondering if anyone had
started hacking around with the official plug-in interface yet. I
included the source to my plug-ins for reference.
Performance Issues:
1. The KLT feature tracking library I have assumes a black and white,
8-bits-per-pixel input image with rowBytes = pixelsWide. Is there any
way to, instead of copying bytes around to ensure the correct input,
simply specify that that's what I need and use the input buffer
directly?
2. Is it less expensive, if I need two subsequent "frames" from the
input stream, to use an accumulator latch in patches, or to snag-and-
buffer two shots within the plug-in? Probably the latter-- if so, how
do I know the input image has changed?
3. In my other plug-in, I try to plot the output from the first plug-
in. I basically turn the image into bytes, malloc a new image of the
same size, copy the old image over, change a few pixels, then output
the new image. This is probably extremely sub-optimal. Ideas?
4. Related to 3, Is valloc better even when I'm changing the data?
--Sam
_______________________________________________
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]
