> Hi All,
> 
> I want to be able to take the image from a camera or similar, and scan it
> to find the edges etc. I also need to be able to pick up similar coloured
> areas inside the outer perimeter.
>

I would read the image into a gWorld...

Then figure out how to put an individual vertical or horizontal line into a
handle.

In a 32-bit Gworld, theroetically a black line would be all zeros (0,0,0,0)
and a white line all 255's (255,255,255,255)

Problem :scanner/digicam -generated images have lots of noise, especially in
the Blue channel.

Problem: how to deal with skewed pure color frames....

By examining the average pixel values in a line and comparing those values
to the next line you should be able to detect a jump from a pure color to
image data.

I'm no mathematician, but the other line property to investigate/compare
would be how large is the largest deviation from the mean pixel value in a
line... Pure color would have small or no deviations from the mean whereas
inage data would have larger deviations....?

(Problem: what about an image that contain pure colors along its edges?)

An interesting problem but a lot of work to do well...

I seem to remember that Andy Gariepy once mentioned that he has code to
detect image edges. Maybe if you ask nicely he will cough it up. (Aside to
Andy: If you feel like coughing, cough my way too....)

Cheers,








----------------------------------------------------------
Michael Evans
Manager of Software Development * Photo Systems, Inc.
3301 Wood Valley Road, NW * Atlanta, GA, 30327-1515
Voice: (404) 846-9386
Fax: (404) 240-0878 * Cell: (404) 229-3930
E-mail: evans@... * michael_evans@...
----------------------------------------------------------