were selected in AVHRR cnanne! 1 and channel! 2. A
related range of brightness level was extracted
from the origina! AVHRR thermal infrared channel 4
according to the ground temperature range except
that of cloud, snow and ice. Sometimes, it was
necessary to adjust the selected leve! range
slightly for some special objective such as bright
desert which has a higher spectral aibedo to cause
a brightness saturation in the lower 8 bits data of
channel 1 and channei 2. The major part of the data
preprocessing was a map projection transformation
of AVHRR data on the basis of Albers equal-area
projection. Satellite orbital parameters and the
time code were used to execute the map projection
transformation. The products of the preprocessing
were placed in file to provide to the data process-
ing of mapping.
DATA PROCESSING OF MAPPING
Data processing of mapping or image base map making
is the most important step in mapping of photomap.
An IS digital image processing system was used to
perform a precise geometric correction, image
enhancement, color adjustment, digital mosaic and
a series of special processings. The image base map
as an output was made by a color fi'm recorder and
a normal darkroom processing.
Precise Geometric Correction
Although the preprocessed images were transformed
into an Albers equal-area projection form, there
were still! a little errors that would cause the
errors of the points position on the finial
photomap to be larger than one millimeter. So it
was necessary to perform a precise geometric
correction to ensure the geometric precision of the
photomap. A published standard map which was
devided into 64 subsections was inputted to the
System as a basis of the geometric control. Using
the general image-register program on the four
quadrants of a subimage, displacement vectors in
each quadrant were determined and a control points
file could be created with these data. Then a
geometric correction was operated with the normal
polynomial correction for every subimages. For some
images, a simple translation was satisfactory
enough to the required precision. The resultant
precision of points position on the photomap was
less than one millimeter under the scale of
1:4,000,000.
Image False Color Composite And Enhancement
In order to get colorful tones, rich information
and the best vision effect, a false color composite
which included a thermal infrared band was taken.
This composite scheme may be the best one in the
composites of AVHRR data, although the degree of
difficulty of color adjustment in a image mosaic is
increased extremely. With the three channels which
are the least correlation in a spectral space,
thermal infrared (assigned as red), near infrared
(green) and visible red (blue), the best „color
balance and the richest information can be achieved.
Before the composite action, the original channel 4
data was processed by a negative operation so that
the image brightness was in proportion to the
ground radiant temperature. As a result, the higher
the radiant temperature is, the bright the red
Color. is. Therefore, the color rule in the
composite fits the human subjective vision habit
psychologically. A controlled linear stretching for
the three components respectively, other than a
Scale based on the statistics of a subimage, was
used so as to contro! the image color for mosaic
and keep a appropriate image contrast and a color
balance.
Digital Mosaic Operation
A high quality and satisfactory digital mosaic
is a
difficult, time-consuming and experience-needed
thing. in this digita! mosaic operation, there were
64 subimages to be mosaic together in the main
part of China and 12 subimages in South China Sea,
We classified the total 76 subimages into © parts
according to the similarity of geographic landscape
and image characteristic, Xinjiang area,Tibet area,
Centrai China, South China, Northeast China and
South China Sea. Mosaic was operated first in these
parts with about 12 to 16 subimages for each one.
Finially, a compeleted mosaic of the 64 subimages
with the 5 parts except South China Sea was taken
to form a embryonic form of the image base map.
Color adjustment is a careful work. Aithough the
histogram matching technique usually is useful in
digital mosaic, It Js just satisfactory on the
condition of the similar geographic landscape and
image character, Therefore, this method is
effective for a limited and related harmonious
area. In our mosaic procedure, histogram matching
just effected in the overlap part of the adjacent
images as a color reference and a interactive color
adjustment was used in the other part of a image.
A feather transition was used to smooth the color
change and remove any artificia! marks.
Some Special Processing Procedures
Cloud Removing Operation: With a series of multi
-temporal and registered images, a cloud removing
operation can be done. Two kinds of image detection
algorithm were needed to extract cloud and cloud-
shadow as the Range Of Interest (ROI) information
to be stored. Cloud detection based on the lower
temperature or lower brightness in channel 4 and
higher brightness in channel 1 and channel 2
Simultaneous!y. The shadow detection was only from
the lower brightness in channel 1 and channe! 2
simu!taneous!y. These can be expressed as a logical
relation as below;
CLOUD = (LOWCHA). AND. (HIGHCH1).AND.(H!GHCH2) — (1)
SHADOW = (LOWCH1}. AND. (LOWCH2) (23
(CLOUD ).OR.( SHADOW) = @ (3)
The CLOUD and SHADOW are the detected area of c!oud
and cloud-shadow in a image respectively. The
LOWCHN and HIGHCHN are the lower and higher bright-
ness of channel n in a image respectively. Both the
detections were needed a proper threshold from a
image. Cloud and shadow removing were done by a
minimum and a maximum operation of multi-temporal
images for the detected cloud and shadow area
respectively. After the operation above, adjust the
brightness of the substituted area slightiy to
eliminate tne difference in brightness between the
substituted parts and the background. A interactive
adjustment should be used for some area if
necessary.
Modification of Image Seasonal Aspect: It s
well known that one of the most important facters
which effects Aa image color directely is seasonal
change of vegetation. Because the images were
received duing a period of neariy 4 months, the
Seasona! change of vegetation in some images was
obvious and couid cause a probiem in the mosaic
especialiy in Northwest China. So, a modification
must be done. Using the vegetation index, such as
CCHO-CHIJ/CCH24CH1) or (CCHOZCHT). we got a
vegetation index image (VI) Then a simple
arithmetic operation was used with the near
infrared data (CH2) and the VI images to form a new
image to substitute the origina! CH2 image.
NEWNIR = À * (NIR) + B x (VID (49
BI A A D> 11)
NIR is a near infrared image (CH2). A and B are
382
