Is there such thing as a 'soft checksum' tool?

[Michaël Grünewald]

perryh at pluto.rain.com wrote:
> Mel Flynn <mel.flynn+fbsd.questions at mailing.thruhere.net> wrote:
> 
> However, thinking about this inquiry and JPEG in the same sentence
> has given me an idea that might help the OP:  JPEG is a "lossy"
> compression, with the degree of loss related to the chosen image
> quality, so two "similar" images might become identical -- or at
> least more similar -- if compressed to a sufficiently low quality
> using the JPEG algorithm.

This seems to be an excellent idea. A similar approach can 
(successfully) be used to let a computer recognize songs through a 
micrOphone: the incoming signal is transformmed to MP3 at a rather low 
quality, which provides a sort of fingerprint of the input. The quality 
factors shall be adujsted adequately for this application: there is a 
tradeoff between stability (noise insensitivity) and separation to find.

The case of images is much more complicated if one wishes to recognize 
the same image at two different scales.
-- 
Cheers,
Michaël