Two-Decoding-Option Image Sharing Method - Visual Cryptography and Secret Image Sharing - page 472

send the raw data (use one bit "1" (or "0") denoting the black (or white)
color) to transmit the black-and-white shadows for Jin et al.'s scheme and
Lin and Lin's scheme. For example, when using the bitmap le format, the
number of bits per pixel, which is the color depth of the image, typically can
be 1, 4, 8, 16, 24, and 32. Even though both schemes use the color depth 1,
our TiOISSS (use the color depth 8) still has the lesser file sizes of shadows for
all cases in Table 17.1. For example, the le sizes of shadows for Jin et al.'s (3,
6)-TiOISSS and Lin and Lin's (3, 6)-TiOISSS are 23592960 bits (23.6 Mbits)
and 5258608 bits (5.3 Mbits) when using one bit to represent the color, while
our (3, 6)-TiOISSS just needs only 1398101 bits (1.4 M Kbits) when using 8
bits to represent the color to transmit a gray-and-white shadow. Notice that
when all schemes are using the same file format of the color depth 8, Jin
et al.'s (3, 6)-TiOISSS and Lin and Lin's (3, 6)-TiOISSS require 188743680
bits (188.7 Mbits) and 42068864 bits (42.1 Mbits), respectively, which are
significantly larger than 1.4Mbits.
By the contrast denition ((hl)/(m + l)) of VCS in [3], the contrasts of
the previewed images are calculated and shown. Since the previewed image is
used for verification in a distributed multimedia system; therefore, our primary
task is to choose the suitable m and g to obtain the reduced shadow size to
make the transmission and storage more ecient.
To elucidate on the good results of the proposed scheme among the exist-
ing TiOISSSs, some properties are evaluated: (1) the resolution of the recon-
structed image, (2) the decoding complexity, (3) the shadow images (including
the image pattern, the shadow size, the file size of shadow), and (4) the pixel
expansion. A comparison among three TiOISSSs in [6, 9, 20] is listed in Ta-
ble 17.2. All three schemes have dual modes of decoding|one for preview (or
when a computer is temporarily unavailable) and the other for perfectly recon-
structing the original gray-level secret image. Also, they all have a distinctive
stacking-to-see property. Our scheme and Lin and Lin's scheme use Lagrange
interpolation for perfect reconstruction with the computational complexity
O(k) and no additional memory space, while Jin et al.'s scheme uses a 8 8
ROM lookup table for reconstruction without computation. Our gray-and-
white shadows have the benefit of reducing the physical size and the file size
of the shadow when compared with other two TiOISSSs.
17.6 Conclusion
Our TiOISSS is a hybrid|half is the VCS and half is PISSS|with each spe-
cialties employed: the easy decoding of the VCS in Phase 1 and the perfect
reconstruction of PISSS in Phase 2. Our combination of the VCS and PISSS
is dierent from that in Lin and Lin's TiOISSS [9]. Lin and Lin use the com-
binations in the m-subpixel block to represent the output value of PISSS,