John Rudd wrote:
No, 70 would still be 70. 07 would become 00. And 07 is a pretty
faint red. Looking at it now, I can't distinguish it from black.
(70 is 0111 so the lower 3 or 4 bits are already 0's, whereas 07
is 0111 .. THAT becomes 0 and is indistinguishable from
black
On Wed, 2 Aug 2006 [EMAIL PROTECTED] wrote:
> Maybe I'm not getting the obvious, but what about using something like
> Perl::Magick to convert a given image into B/W? I mean, ImageMagick is
> made for things like that...
> Shrinking it to, say, a quarter of it's original size would take care of
>
>
> On Aug 2, 2006, at 3:03 AM, Matthias Keller wrote:
>
will it not be much faster just to make a md5 sum on the image file
without
thinking if it a appel or orange ? :-)
>>> Yes, but just taking a straight sum will be sensitive to all of those
>>> small pixels which are changed by
On Aug 2, 2006, at 3:03 AM, Matthias Keller wrote:
will it not be much faster just to make a md5 sum on the image file
without
thinking if it a appel or orange ? :-)
Yes, but just taking a straight sum will be sensitive to all of those
small pixels which are changed by the spammers so that th
will it not be much faster just to make a md5 sum on the image file
without
thinking if it a appel or orange ? :-)
Yes, but just taking a straight sum will be sensitive to all of those
small pixels which are changed by the spammers so that they have
different sums, but the differences aren'
On Aug 2, 2006, at 12:12 AM, Benny Pedersen wrote:
On Wed, August 2, 2006 06:11, John Rudd wrote:
white will produce (assuming 24bit color) f0,f0,f0 and black will
produce 00,00,00. Thus, you get a nice high-contrast image for
feeding
just for clearness white is ff, ff, ff
yes, white i
On Aug 1, 2006, at 11:58 PM, Derek Harding wrote:
John Rudd wrote:
On Aug 1, 2006, at 10:30 PM, Derek Harding wrote:
John Rudd wrote:
Um, how exactly will they fail?
How about a nice black & white speckled image with red text on it?
Explain to me how you think it will fail?
So you're
On Wed, August 2, 2006 06:11, John Rudd wrote:
> white will produce (assuming 24bit color) f0,f0,f0 and black will
> produce 00,00,00. Thus, you get a nice high-contrast image for feeding
just for clearness white is ff, ff, ff
will it not be much faster just to make a md5 sum on the image file
John Rudd wrote:
On Aug 1, 2006, at 10:30 PM, Derek Harding wrote:
John Rudd wrote:
Um, how exactly will they fail?
How about a nice black & white speckled image with red text on it?
Explain to me how you think it will fail?
So you're dropping three bits? White is FF, Black 00,
On Aug 1, 2006, at 10:30 PM, Derek Harding wrote:
John Rudd wrote:
Um, how exactly will they fail?
How about a nice black & white speckled image with red text on it?
Explain to me how you think it will fail?
John Rudd wrote:
Um, how exactly will they fail?
How about a nice black & white speckled image with red text on it?
BTW I think the OCR approach is unlikely to succeed due to processing
constraints.
Derek
On Aug 1, 2006, at 8:55 PM, Loren Wilton wrote:
2) to combat the "images with subtle differences", develop a checksum
method that ignores the lower (3 or 4 bits? out of 8 bits) of each
color channel. That way you get what is essentially a very high
Won't work. White on black and black on w
2) to combat the "images with subtle differences", develop a checksum
method that ignores the lower (3 or 4 bits? out of 8 bits) of each color
channel. That way you get what is essentially a very high
Won't work. White on black and black on white are both quite readable, and
will fail the ab
1) use Martin Blapp's OCR plugin/patch for SA. feed data to bayes.
http://antispam.imp.ch/patches/patch-ocrtext
2) to combat the "images with subtle differences", develop a checksum
method that ignores the lower (3 or 4 bits? out of 8 bits) of each
color channel. That way you get what is e
14 matches
Mail list logo
