image content scanning

Glenn Steen glenn.steen at
Wed Aug 1 08:54:53 IST 2007

On 31/07/07, Brent Addis <brent.addis at> wrote:
> yeah I do, its pretty average at what we need. It does catch some, but not some of the more hardcore stuff.
> ________________________________
> From: mailscanner-bounces at on behalf of Glenn Steen
> Sent: Tue 31/07/2007 11:01 p.m.
> To: MailScanner discussion
> Subject: Re: image content scanning
> On 31/07/07, Brent Addis <brent.addis at> wrote:
> > ?
> > Hi,
> >
> > Has anyone found anything useful for looking at porn images in email, not
> > just from a spam perspective, but from a general content perspective.
> > Something that checks skin tone for example. I am not really concerned how
> > many cpu cycles that this sort of thing consumes. Is there a plugin for
> > spamassassin or mailscanner that isn't widely known about that will do this?
> >
> > Thanks,
> >
> > Brent
> You already use ImageInfo (, I presume?

Well, the problem with "skin tones detection" is that it is pretty
error-prone in more ways than one...
- Ethnicity... Plays havoc with this
- FP on normal images... Think head shots
- FP/FN due to the non-trivial nature of detecting/sampling the images
correctly (simply looking on the "bias" of the colour gamut (could be
a simple histogram) for that image will not do more than a shoddy
work... Will be light on resources though very error prone, IMO).
- addendum to the last bit is that it would likely be rather resource hungry.
I used to work at a firm making ID cards/drivers licenses (had a
monopoly, more or less, back in the early 90's, in Sweden)... We had a
rather nifty image capture/"improvement" system, and that had a hard
time (well...) finding such simple things as a "white" background
(they never are white:-)... And then we didn't have that much
variation wrt ethnicity...;).

So, in short, what seems like a good idea isn't exactly the best
possible one... Might be why, to my knowledge, there is no such thing
implemented... yet.

-- Glenn
email: glenn < dot > steen < at > gmail < dot > com
work: glenn < dot > steen < at > ap1 < dot > se

More information about the MailScanner mailing list