Hello Sarah, hello all,
some times ago I had the same problem and came accross this blogpost 
that gives some advice about how to developp an algorithm that would do 
that : 
http://www.hackerfactor.com/blog/index.php?/archives/432-Looks-Like-It.html 
. The update section on this blogpost has some links to scripts that 
implement this algorithm. My project was canceled so I can't tell you if 
it works well but according to comments it can be a good solution. But 
if you have money to spend on it, http://services.tineye.com/MatchEngine 
is definitely a good choice as their algorithm is quite good.

Hope this helps (and maybe that since my original research is quite old 
now, there are now packaged tools that do it, I don't know)

--
Sylvain Machefert - Web services librarian
http://geobib.fr/en




Le 16/10/2014 18:38, Kyle Banerjee a écrit :
> Could you say something about the type of dup detection you need? Are we
> talking true duplicates, or possibly the same image in multiple formats,
> cropped, etc? Roughly how many images (thousands, tens of thousands, etc)
> and how big are they? Also, what did you try that did not meet your needs?
> Thanks,
>
> kyle
>
> On Wed, Oct 15, 2014 at 2:56 PM, Shipley, Sarah <[log in to unmask]>
> wrote:
>
>> Hi,
>>
>> I was wondering if anyone had any recommendations for image de-duping
>> software that compares the images rather than checksums.  We're using
>> Visual Similarity Duplicate Image Finder, but find it's not as accurate as
>> we'd like.   We have a very large number of images to de-dupe in our photo
>> archives and with the current software can't find a balance of comparison
>> that finds all the dups without producing a lot of false positives.
>>
>>
>> [cid:[log in to unmask]]
>>
>> Sarah Shipley, CA
>> Digital Asset Manager
>> Legislative Department - Office of the City Clerk
>> http://www.seattle.gov/leg/clerk/
>> 600 Fourth Avenue, Floor 3
>> PO Box 94728
>> Seattle, WA 98124-4728
>> 206.684.8119
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>
>>