Image Secret Sharing - Visual Cryptography and Secret Image Sharing - page 378

to generate visual cryptograms for color images [19]. Using this color decom-
position, every color within the image can be decomposed into one of three
primary colors: cyan, magenta, or yellow. This proposal is similar to tradi-
tional visual cryptography with respect to the pixel expansion that occurs.
One pixel is expanded into a 2 2 block where two color pixels are stored
along with two transparent (white) pixels.
However, [21] examined the security of Hou's [19] scheme, and while the
scheme is secure for a few specific two-color secret images, the security cannot
be guaranteed for many other cases.
Improving this pixel expansion and also working out the optimal contrast
of color visual cryptography schemes have been investigated [10]. In the paper,
they prove that contrast-optimal schemes are available for color visual cryp-
tography (VC) and then further go on to prove the optimality with regard to
pixel expansion.
A lossless recovery scheme outlined by [20] considers halftoning techniques
for the recovery of color images within visual cryptography. The scheme gen-
erates high quality halftone shares that provide lossless recovery of the secrets
and reduces the overall noise in the shares without any computational com-
plexity. Their proposed method starts by splitting the color channels into its
constituent parts, cyan (C), magenta (M), and yellow (Y). Each channel has
grayscale halftoning applied to it. Error diffusion techniques discussed in [35]
are then applied to each halftone channel. A circularly symmetric filter is used
along with a Gaussian filter. This provides an adequate structure for the dot
placement when constructing the shares.
Eciency within color visual cryptography [25] is also considered which
improves on the work done by [34, 3]. The proposed scheme follows Yang and
Laih's color model. The model considers the human visual system's eect on
color combinations out of a set of color subpixels. This means that the set of
stacked color subpixels would look like a specific color in original secret image.
As with many other visual cryptography schemes, pixel expansion is an issue.
However Shyu's scheme has a pixel expansion of dlog 2 ce, which is superior to
many other color visual cryptography schemes especially when c, the number
of colors in the secret image becomes large. An area for improvement however
would be in the examination of the difference between the reconstructed color
pixels and the original secret pixels. Having high quality color VC shares would
further improve on the current schemes examined within this survey, this
includes adding a lot of potential for visual authentication and identification.
Chang et al. [6] present a scheme based on smaller shadow images, which
allows color image reconstruction when any authorized k shadow images are
stacked together using their proposed revealing process. This improves on the
following work [32], which presents a scheme that reduces the shadow size
by half. Chang et al.'s technique improves on the size of the share in that,
as more shares are generated for sharing purposes, the overall size of those
shares decreases.
In contrast to color decomposition, Yang and Chen [33] propose an addi-