On Tuesday 23 March 2010, Aditya Bhatt wrote:
> @K?re
>
> Just a comment about the image size tradeoff.
>
> > I have just updated the automatic image selection on previews in Skanlite
> > (actually libksane). What I did was to first do a rough auto selection on
> > a resized image ~100 * 150 pixels and then refine the selections on the
> > full- sized preview. It improved the speed dramatically and did not
> > decrease the accuracy. (actually it removed a lot of false positives)
> >
> > Just an idea if you needed more :)
>
> Hi, what are you detecting in the image? What are the selections that you
> talk of?
> ( I can't checkout code right now, since the firewall in my college is
> something of a joke ATM, it blocks svn://)
> I'm not entirely sure what you mean, but thanks anyway :)
I have a really simple problem compared to recognizing faces :) I try to
detect the square areas around pictures in the preview on a flatbed image
scanner.
>
> Actually what I meant was that when I run multiple cascades (5 or 6) , the
> process is very slow. It would be faster if the image is smaller, but
> there's of course a limit on how small an image I can make. The sad part
> is, I can't resize it to 100*150 pixels and expect good detection
> accuracy, because there are supposed to be a lot of faces in a photo and
> such a tiny image size will cause a phenomenal loss of information :)
>
Yep, as you and Michael noted 100*150 pixels is not enough for a group photo.
The situation is a bit different for the scanner. There you probably do not
want to have more than ~5 different selection and there a small resolution
image is enough.
The main idea could still be valid. The idea being to have a "two phase
detection". The first phase would does a very quick inaccurate search (false
positives?) and the second would only investigate the areas found in the first
phase.
> If you can help here, it'd be great :)
I'm afraid I would not be of much help ;)
Kåre
_______________________________________________
Digikam-devel mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-devel
Locked
