Graphics Reference
In-Depth Information
In this research paper, we introduce a method which allows the use of the remote
sensing image's gray-scale characteristic to extract quickly the suspected targets. The
method aims to reduce human consumption in the process of interpretation and, at the
same time, alleviate the pressure of the photographic interpreter training.
The rest of this essay is arranged in the following way: the second part introduces
image pretreatment; the third part introduces the suspected target extraction method;
the fourth part gives the experimental and analysis results; and finally, the fifth part is
the conclusion.
2
Image Pretreatment
2.1
Reducing Image Resolution in One Process
Because the remote sensing image has high resolution, it requires not only a huge
amount of storage, but also a reduced speed processing. Reducing the resolution of
the processed image can help to solve above mentioned problems. For this purpose,
two solutions are put forward in the following part.
(1) Resolution Reduction
f
(
a
,
b
)
A ×
B
If the size of the original image
is
, compressing the original image to
g
(
a
N b
,
)
A
×
B
compress the image
, which size is
, proportionally. N, equals
N
N
N
A
B
to
(also means
), is the size of the compression window. Reducing pixels in
A
B
g
(
x
,
y
)
the window to one pixel point, which gray value
is the average value of
N
N
those pixels in the window.
f
(
x
,
y
)
( 1 )
(
x
,
y
)
M
g
(
x
,
y
)
=
N
N
N
*
N
N is the size of the template. M is the set of pixels in the template. The original im-
age's pixel coordinate
(
x
,
y
)
corresponding to the compressed image's pixel coordi-
x
y
(
)
x
,
y
=
(
,
)
nate
.
N
N
N
N
Since the image compression reduces the size of the original image, the amount of
image data is naturally decreased; and consequently this contributes to consume mi-
nor storage capacity and improve the processing speed.
(2) Image Segmentation
N, the number of blocks after the segmentation, is determined by the size of image. In
order to ensure that one extracted target exists completely in at least one image after
Search WWH ::




Custom Search