CLOUD REMOVAL OF OPTICAL IMAGE USING SAR DATA FOR ALOS
APPLICATIONS.
EXPERIMENTING ON SIMULATED ALOS DATA
Nguyen Thanh Hoan*, Ryutaro Tateishi
Centre for Environmental Remote Sensing (CEReS), Chiba University, 1-33 Yayoi-cho, Inage-ku, Chiba, 263-6522,
Japan-nthoan)@graduate.chiba-u.jp, tateishi@faculty.chiba-u.jp
Commission IV, WG IV/3
KEY WORDS: Remote Sensing, Identification, Point Cloud, Radar, Software, Orthoimage, Simulation, Integration
ABSTRACT:
Cloud is always problem of optical remote sensing data. ALOS-AVNIR2 images also will be affected. Microwave images are not
affected by cloud. ALOS satellite has both optical and microwave sensor, similar resolution (~10m). Therefore, main purpose of this
study is to develop a method for removing cloud in ALOS-AVNIR2 images based on ALOS-PALSAR data. This method is
developed based on interpolating under cloud pixel values for ALOS-AVNIR2 images. To remove cloud, it needs to be defined. A
combination method of Total Reflectance Radiance Index (TRRI) and Cloud-Soil Index (CSI) is used to define cloud. Around cloud
pixels are mixture of cloud and other objects. So, that is very difficult to define. Therefore, the around cloud pixels are extended
from cloud. Cloud shadow problem also is discussed in this study. Condition to apply this method is that objects in optical image
and radar image change not very much. This method is experimented on simulated ALOS data from Landsat TM and JERS1-SAR
images. Interpolated image is a free cloud and shadow image. Visual logic of objects is good. Original image and interpolated image
is almost similar together. This method also can be applied for combination of two optical images to remove cloud if change of
objects covered by cloud is not so much. Result of this study is a program with many functions like: define cloud, extend cloud, get
shadow, remove cloud and so on. This is free software for every user.
1. INTRODUCTION
Cloud is always problem of optical remote sensing data. And of
course, ALOS-AVNIR images will be also affected. ALOS-
PALSAR data is not affected by cloud. This study proposes a
method to remove cloud in optical images based on
interpolation from SAR data. Combination of Total Reflectance
Radiance Index (TRRI) and Cloud-Soil Index (CSI) is used to
define cloud. TRRI is an index developed by Prof. Nguyen
Dinh Duong, VAST, Vietnam (N.T.Duong, 1998) and CSI is an
index developed by the author for this study. Some other
technics are used to refine cloud like remove single pixels,
extend cloud and so on. Shadow of cloud is also solved by
interpolating from cloud based on direction and distance from
cloud to shadow (these problems will be explained in the
Methodology section). Requirement of this method is that
objects in ALOS-AVNIR2 and ALOS-PALSAR data are not
different very much. So, this method can apply for the data in
which optical image and microwave image in the same
geographic place are not very far together about time. With
ALOS satellite, hope that this requirement will be satisfied.
This method also can be applied for removing cloud from
optical images based on other optical images. In this study,
some examples of removing cloud in ASTER image based on
Landsat-ETM image, removing cloud in Landsat-TM image
based on Landsat-TM image is also shown. This method can
help to make a series of free cloud multi-temporal images for
change detection studies, environmental monitoring studies and
so on.
2. SIMULATED ALOS DATA
This study uses simulated ALOS data. ALOS-AVNIR2 has 4
spectral channels. Wavelength of them is close to Landsat TM.
Comparison table is shown as following:
Band
ALOS - AVNIR2
Landsat TM
1
420 - 500 nm
450 - 520 nm
2
520 - 600 nm
530 - 610 nm
3
610 - 690 nm
630 - 690 nm
4
760 - 890 nm
750 - 900 nm
Table 1. Comparison of ALOS-AVNIR2 and Landsat-TM
So, the first 4 bands of Lansat TM are used instead of ALOS-
AVNIR2 image.
Frequency of ALOS-PALSAR is close to JERS1-SAR data as
in the below table. So, JERS1-SAR data is used instead of
ALOS-PALSAR data.
ALOS - PALSAR
JERS-1 SAR
Frequency
1270 MHz
L-band
1300 MHz
L-band
Table 2. Comparison of ALOS-PALSAR and JERS1-SAR