tanbul 2004
ntext) (6)
0 the map’
ly. We can
how much
solution. A
nethod.
n this new
f entropy?
P? which is
‘ed through
purpose of
incertainty”
to both the
s map is of
cannot be
it is useless
tion to the
ontains few
and do not
'er than the
map) can
a higher
tion theory
nation, but
ther all of
t. It is not
ice on how
distributed
juantitative
t A brief
Is begin by
one pixel,
on amount
for we can
egin. Then
r and finer
on amount
nation will
ails can be
| of data 15
ust usually
peak lies
"s display
search and
how much
his what
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B2. Istanbul 2004
4. EXPERIMENTAL EVALUATION
In this section we present a more precise description to our
method and make some experimental research to reveal the
relation between the information amount and the data amount.
4.1 Calculation of raster-based information
To two adjacent pixels in a gray level image, the difference
between them is the absolute value of the subtraction between
their gray values. We can apply the same way to a color image
(Le. RGB) by decomposing it into several components and
dealing with each component separately. So we only consider
gray level images in the following, the method can be applied to
color images in the same way.
As mentioned in Part 3, information of a pixel is defined as the
difference with its context. To evaluate this difference
quantitatively, we proposed a formula to calculate this
information as following. Let P be the pixel we want to
C
calculate, the gray value of P is ^ and it has i adjacent pixels
! Then the information can be calculated as:
(7)
: ; 1,
inf ormation,, , = nA =C
i=l
To a certain map, the information amount is the sum of all
pixels’ information:
inf ormation
= > inf ormation,, (8)
i
Now we make a deeper inspection into the Internet spatial
information service. In this type of service, through the so-
called three-tiered architecture, Web servers receive and
process client request and send responds, application server deal
with business logic and communicates with database via some
APIs to obtain the data clients want. When users are sending
request for a spatial information service, they rarely are
interested in the total map, oppositely, there are often some
regions or layers that they show their interest in, and these
regions or layers keep changing when users are roaming on the
map. As for the server, we can think of this activity as opening
a small window on the map and sending contents covered by
this map to the certain user.
Being clear with what the users do when they are requesting for
spatial information service, we can utilize the definition of
information we proposed before to direct the data transmission
of this spatial information service. See, to a certain map, we
design our experimental scheme as following. The map stored
on the server side is one with high resolution and not being
totally requested by the user. A small window, which is on
behalf of the scope of map requested by the user, will be used to
clip the total map, and what clipped from the total map is the
data that may be transferred to the client (but not all of them
definitely). This small window may be relatively large, when
User want to cover more content on the whole map, even may
be the same with the total server map, and it may be relatively
369
small when user is interested with some details of a local part of
the server map.
Since the possible data that may be transferred is decided by the
small window on the server side through clipping, the next
question is how much of this data must be transferred. To solve
this question the new information definition proposed by us can
be utilized. From the analysis above we know the amount of
this information is relevant to the data amount and the
resolution, this is the metric we need. Suppose user browse the
map with a window with certain size, then the data which is
clipped by the server small window will be displayed in this
window and we use this window to compute the information
amount and the corresponding data amount. The whole process
is illustrated in the following Fig.8.
The Server Small Clipping Window
j ; DS Window
The Whole Map on Server Side / TheClient Browsing Windo
[4
Figure 8. The Process of Spatial Information Service
4.2 Experimental results and evaluation
The Experimental data and results are given as following
Figure.9 and Figure.10. To a certain small server small clipping
window, an experimental result like Figure.10 can be derived to
describe the relationship between data amount and information
amount at this certain server small clipping window size.
E. y^ a
Figure 9. The Experimental Map