[Image-SIG] content based image querying

[Daniel Fetchinson]

Hi folks,

I'm aware that this subject is a huge one but please bear with me, I
think I have well defined questions, although I'm not at all an
imaging expert. Math is not a problem though.

I've been looking into content based image querying and was lead to
the following paper:

Jacobs, Finkelstein, Salesin: Fast Multiresolution Image Querying -
http://salesin.cs.washington.edu/abstracts.html#MultiresQuery

This paper has the following appealing characteristics:

1. simple to understand the algorithm for non imaging experts
2. relatively easy to implement
3. what it does is exactly what I need

The paper is from 1995 and I was wondering if in the past 13 years
there was any comparable paper with the above 3 characteristics ("what
I need" can be defined as "what the above paper does").

Also, experts in the imaging field also think that this paper is a
relevant one? Are there better approaches to the same problem? (Again,
"same" means what the above paper does.)

Concerning implementation I'm not afraid of writing C but would prefer
something that can be implemented using PIL or stable third party
libraries like gegl. For example the Haar transform always pops up in
similar papers and I'm not aware of a stable library that would do
this for me. Are there any I should look into?

Finally, I'm aware of the imgSeek project but its not suitable for my
needs for various reasons, dependency on qt, slow development, weird
database approach, weird packaging (imgSeek, imgSeekLite, isk-daemon,
pretty chaotic), not cross-platform (32bit vs 64bit issues), etc. I
really considered the option of using parts of imgSeek but after a
careful deliberation I decided to start from scratch.

Cheers,
Daniel

-- 
Psss, psss, put it down! - http://www.cafepress.com/putitdown