Re: My thoughts on image spam strategies

[Loren Wilton]

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 above test.

The test that needs to be made is to measure the visual difference between 
adjacent color patches and assign a weight based on how easy or difficult it 
is for the AVERAGE human eye to detect the color/density difference.  Ie: if 
this is the edge of the text and background, can you tell?

This is harder to do than your test, but there are algorithms that can do 
exactly this.  Whether they would be cost effective is another question. 
And how easily they can be bypassed.

       Loren