Re: [help-GIFT] Searching for similarities

[David Squire]

Wolfgang Mueller wrote:

> David "Viper features" McG. Squire might have something to say to that?

Absolutely. The current Viper features were always something of a hack - though 
they have lasted as long as they have because they work surprisingly well (at 
least versus our initial expectations :) ).

The texture computation is what takes time, and may well not be so relevant for 
this "near duplicate" detection. Setting up an "algorithm" which used only the 
colour features would be much faster (and trivial to do), and may even give 
better results for this task.

Other people have worked specifically on this problem, for example 
http://www-db.stanford.edu/~wangz/project/imcopy/SPIE99/

It is perhaps worth pointing out that I have a new version of CFeedbackClient 
in the pipeline,, and am thinking of doing quite a major overhaul of it to make 
it more robust and flexible (at the moment we have dangerous things like 
functions which accept variable numbers of arguments but in which argument 
order is critical...)

Cheers,

David


--
Dr. David McG. Squire, Postgraduate Research Coordinator (Caulfield)
Computer Science and Software Engineering, Monash University, Australia
http://www.csse.monash.edu.au/~davids/   http://viper.unige.ch/
Do/Don't want HTML mail? Let me know.