Cryptography Reference
In-Depth Information
Figure 9.4: These are the most-significant bits of the photo in Fig-
ure 9.1. The seven least significant bits were deleted to contrast the
images in Figure 9.2.
Nor is there any reason why only 1 bit is encoded in each pixel.
The same algorithms that choose 256 or 128 best colors for an im-
age can be used to find the closest 64 colors. Then 2 bits per pixel
could be allocated to hidden information. Obviously this can lead
to a degraded image, but the hidden information can often mod-
erate the amount of degradation. Imagine, for instance, that we
tried to be greedy and hide 4 bits per pixel. This would leave 4
bits left over for actually specifying the color and there could only
be 16 truly different colors in the table. If the photo was of a
person, then there is a good chance that one of the colors would
be allocated to the green in the background, one of colors would
go to a brown in the hair and maybe two colors would be given
over to the skin color. A two-toned skin could look very fake.
2
But each of these two tones might also be hiding 4 bits of informa-
tion. This would be mean that there were 16 surrogates for each of
these two tones and these 16 surrogates would be used fairly ran-
2
Recent work suggests that human eyes pick up skin tones more than most colors.
So the best algorithms devote more colors in the table to skin colors in the hopes of
better representing them. The eyes don't really seem to care much about the shade of
green in a tree.
