From: "ara.t.howard" <ara.t.howard@...>
Date: 2008-06-24T00:04:31+09:00
Subject: Re: Texture/image similarity with ruby? (computer vision)


On Jun 19, 2008, at 5:33 AM, Casimir wrote:

> Still no luck so far, still looking. Anyone?

a good toolkit:

http://kogs-www.informatik.uni-hamburg.de/~koethe/vigra/

and another

http://www.itk.org/HTML/WatershedSegmentationExample.html



i've used bot a lot, shelling out to ruby mostly, unfortunately  
require hacking in c--.  insight has ruby bindings though (via swig).


i think you are going to want to do some filtering in the frequency  
domain focusing on edge images (shapes) and colours, possibly warping  
to achieve a best fit.  whatever you do is going to be entirely custom  
- as everything in computer visions tends to be, despite having a  
plethora of off the shelf tools to start from.

kind regards.

ps.

i nearly always end up stuff pixels into narrays and doing some manip  
there - if nothing else to refine algorithms in ruby before  
laboriously re-coding in c-- or c.  for this is useful to have  
representatvie images that are small and as close to raw as possible.



a @ http://codeforpeople.com/
--
we can deny everything, except that we have the possibility of being  
better. simply reflect on that.
h.h. the 14th dalai lama