The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
380
In this study, Landsat TM image date 1992/10/21 and JERS1-
SAR image date 1992/10/27 of Hanoi, Vietnam are used.
JERS1-SAR image was resampled to 30m resolution to
correspond with Landsat ETM image. It also is applied
GAMMA 3X3 FILTER to reduce noises. And it was corrected
about geometry by orthogonal geometric correction before
using. Following figure is JERS1-SAR image after these
processing:
By this way, we have perfect data for experimenting the remove
cloud method including: cloud image to test the cloud defining
method, to test the interpolation method and we also have
original data (under cloud data - before overlay cloud) to
compare with interpolated result for validation. Following is
Landsat-TM image after overlaying cloud
Figure 3. Landsat TM image date 1992/10/21 after overlaying
cloud
Figure 1. JERS1-SAR image 30m resolution after GAMMA 3. METHODOLOGY
3X3 FILTER and Orthogonal Geometric Correction
Outline of methodology is illustrated in the following figure:
In the Landsat TM image, there is no cloud. Cloud was gotten
from another Landsat TM image and overlay to Landsat TM
image date 1992/10/21 as following figures: Methodology
ALOS-AVNIR2 Image
Define Cloud
ALPS PALSAR data
ÄfflSÄCoiKctipn
JÄeSSdl
Get Shadow
Tmm
reduce*
INTERPOLATION
Figure 4. Flow chart of methodology
3.1 Defining Cloud
Reflectance of cloud is always high from visual wave to near
infrared wave. So, values of cloud will be high in all 4 bands of
ALOS-AVNIR data. With thick cloud, we are very easy to
define it by TRRI index. This index is defined by the following
formula:
TRRI
J ^ max A
* 100
Figure 2. Landsat TM image date 1992/10/21 (up) and Landsat
TM image used to get cloud (down)