Bug #56118 for Imager: Filter type gaussian has problems for large values of std dev

Tue Mar 30 13:57:26 2010 johnrcomeau [...] gmail.com - Ticket created

Subject:

Filter type gaussian has problems for large values of std dev

I'm using Perl version 5.10.0 on Ubuntu Linux. and Imager version 0.73 Large values of standard deviation don't work for $image->filter(type => 'gaussian', stddev = > $stddev). The image returned is gray. The maximum working stddev has sometimes been 6 and other times 20. Maybe it's dependent on the image? In any case, there is an upper limit to the standard deviation, and how is one to know that, and shouldn't an error string be available when the limit is reached? As a workaround, can one filter the same image N times to achieve a standard deviation sqrt(N) times greater? -John Thanks, John

Wed Mar 31 06:47:30 2010 TONYC [...] cpan.org - Correspondence added

On Tue Mar 30 13:57:26 2010, johnrcomeau wrote: Show quoted text

> I'm using Perl version 5.10.0 on Ubuntu Linux. > and Imager version 0.73 > > Large values of standard deviation don't work for > $image->filter(type => 'gaussian', stddev = > $stddev). > > The image returned is gray. > > The maximum working stddev has sometimes been 6 and other times 20. > Maybe it's dependent on the image? In any case, there is an upper limit > to the standard deviation, and how is one to know that, and shouldn't an > error string be available when the limit is reached?

I just tried some images with stddev set to 50 with the expected result - colour was preserved. These were large images (3264 x 2448) from my camera. I'd expect some sort of greying effect of a high strength blur on a lower resolution image simply because so many pixels are being averaged. If you have a particular image that isn't behaving the way you expect, please send me the image and the code you've written using Imager that mis-behaves, and I'll take a closer look at it. Show quoted text

> As a workaround, can one filter the same image N times to achieve a > standard deviation sqrt(N) times greater?

Yes, but I'd expect the same greying effect. Tony

Wed Mar 31 06:47:31 2010 The RT System itself - Status changed from 'new' to 'open'

Mon Apr 12 04:18:02 2010 TONYC [...] cpan.org - Priority changed from (no value) to '40'

Mon Apr 12 04:18:03 2010 TONYC [...] cpan.org - Status changed from 'open' to 'stalled'

Sun Jun 20 05:37:59 2010 TONYC [...] cpan.org - Correspondence added

On Wed Mar 31 06:47:30 2010, TONYC wrote: Show quoted text

> On Tue Mar 30 13:57:26 2010, johnrcomeau wrote:

> > I'm using Perl version 5.10.0 on Ubuntu Linux. > > and Imager version 0.73 > > > > Large values of standard deviation don't work for > > $image->filter(type => 'gaussian', stddev = > $stddev). > > > > The image returned is gray. > > > > The maximum working stddev has sometimes been 6 and other times 20. > > Maybe it's dependent on the image? In any case, there is an upper limit > > to the standard deviation, and how is one to know that, and shouldn't an > > error string be available when the limit is reached?

> > I just tried some images with stddev set to 50 with the expected result > - colour was preserved. > > These were large images (3264 x 2448) from my camera. > > I'd expect some sort of greying effect of a high strength blur on a > lower resolution image simply because so many pixels are being averaged. > > If you have a particular image that isn't behaving the way you expect, > please send me the image and the code you've written using Imager that > mis-behaves, and I'll take a closer look at it. >

> > As a workaround, can one filter the same image N times to achieve a > > standard deviation sqrt(N) times greater?

> > Yes, but I'd expect the same greying effect.

Closing since there was no response and the problem couldn't be reproduced. Tony

Sun Jun 20 05:38:00 2010 The RT System itself - Status changed from 'stalled' to 'open'

Sun Jun 20 05:38:00 2010 TONYC [...] cpan.org - Status changed from 'open' to 'rejected'