PanoTools mailing list archive
|Date/Time:||Mon, 20 Feb 2006 16:45:09 +0100|
|Subject:||Re: Re: PTGui 5.7 - how to use the antialiasing filters ?|
on Monday, February 20, 2006, 15:54:34 you wrote:
>> EK> Could you comment on how large the kernels are?
>> I have added the values to the http://www.pano2qtvr.com/dll_patch.php
>> page. This size is without shrinking.
EK> Do I understand right that the largest kernel is 4x4 pixels? This is far smaller
EK> than half of the "old" interpolators, which used up to 32x32 pixels (sinc1024).
The kernel size on my page is the radius of the kernel so the largest
kernel is 8x8 pixels. Its also possible to extend this size to what
ever you like, but I see no real advantage since the the maximum
contribution of a pixel beyond this radius with a gamma of 2.2 is
sinc(4.5 * Pi) ^ 2.2 = 0,00295 for a 8 bit image this makes 0,754 of a
step... so a really wide pattern of white/black pixels would be needed
to see any difference at all... and I doubt that this gives "better"
results instead of just echos in a normal image.
>> EK> And do you have any estimation about execution times?
>> I would say 50% to 100% slower than the original filters without fast
EK> 100% slower means no speed at all - infinite execution time? ;-)
yes... some times... ;))))
EK> kernel size dependent from downsampling factor means that execution time highly
EK> depends on what you do. One would have to write a true benchmark with an average
EK> usage profile...
so we should take 1000 pts project files of the users of this list...
;) This shows the general problem of benchmarks.... The new Intel iMac
is twice as fast as the old one... and the old was twice as fast as an
Yahoo! Groups Links
<*> To visit your group on the web, go to:
<*> To unsubscribe from this group, send an email to:
<*> Your use of Yahoo! Groups is subject to: