Databases Reference
In-Depth Information
F I GU R E 7 . 5
The pyramid structure for progressive transmission.
Instead of taking the arithmetic average, we could also form some sort of weighted average.
The general procedure would be similar to that described above. (For one of the more well-
known weighted average techniques, see [
95
].)
The representative value does not have to be an average. We could use the pixel values
in the approximation at the lower levels of the pyramid as indices into a lookup table. The
lookup table could be designed to preserve important information such as edges. The problem
with this approach would be the size of the lookup table. If we were using 2
2 blocks of
8-bit values, the lookup table would have 2
32
values, which is too large for most applications.
The size of the table could be reduced if the number of bits per pixel was lower or if, instead
of taking 2
×
2[
96
].
Finally, we do not have to build the pyramid one layer at a time. After sending the lowest-
resolution approximations, we can use some measure of information contained in a block to
decide whether it should be transmitted [
97
]. One possible measure could be the difference
between the largest and smallest intensity values in the block. Another might be to look at
the maximum number of similar pixels in a block. Using an information measure to guide the
progressive transmission of images allows the user to see portions of the image first that are
visually more significant.
×
2 blocks, we used rectangular blocks of size 2
×
1 and 1
×
7.7 Facsimile Encoding
One of the earliest applications of lossless compression in the modern era has been the com-
pression of facsimile, or fax. In facsimile transmission, a page is scanned and converted into
a sequence of black or white pixels. The requirements of how fast the facsimile of an A4
document (210
×
297 mm) must be transmitted have changed over the last two decades. The
