The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008
383
of them in interpolated image is not clear. This problem is
Figure 12: Problem in the interpolated image (up-left), Cloud
image (up-right), Original optical image (down-left) and radar
image (down-right)
This object was checked in radar image but that cannot be
separated by radar image from around object. This problem
may be because of radar image (reflectance is the same with
around objects) or change of object (radar image and optical
image are different 6 days). This checking is illustrated in the
figure 12.
These objects were estimated by local farmers. That is one type
of vegetable. Height of this vegetable is about 20-30cm. So,
reflectance in radar image may be from ground, not from
vegetable. Cause of radar image (can not separate this object by
radar image) is higher probability than cause of change of
object. So, interpolation of under cloud data based on radar
image will be not so good at some special objects. However, by
using combination of threshold level (a value) for radar image
and searching the nearest pixel to interpolate cloud pixel, the
result will have more information than using threshold of whole
radar image and cloud optical image independently.
4.2 Cloud Removal of ASTER Image Based on Landsat
ETM Image
This method also can apply for combination of two optical
images to remove cloud if the change of objects is not so much.
By the same way, cloud was gotten from another image to
overlay on ASTER image date 2001/11/16. NDVI of Landsat-
ETM image date 2000/11/04 was used instead of radar image.
shown in the figure 12.
Interpolated image is compared with original image by visually.
Result is very good. Almost there are no change between
interpolated result and original ASTER image. Result is shown
in the following figure:
Figure 13: Original image (left), Cloud image (centre) and
Interpolated image (right) of ASTER image date 2001/11/16
By this way, many types of different optical image can combine
together to remove cloud. We can make a series of free cloud
multi-temporal data from many types of satellite data for
studies about monitoring the change of environment and natural
resource.
4.3 Combination of Landsat-ETM Image and Landsat-
ETM Image
Here is another application of this method. Two Landsat-ETM
images date 2002/01/05 and 2001/02/03 were used. Both of
them have cloud. These two images can be combined together
to make two free cloud images. In this case, wavelength of two
images is the same together. So, interpolation is done based on
band by band. That means using band 1 to interpolate for band
1, band 2 to interpolate for band 2, band 3 to interpolated for
band 3 and so on. The first, cloud part in the image date
2001/02/03 was removed based on the image date 2002/01/05.
Result is shown in the following figure:
Figure 14: Cloud image (left) and resuit image (right) of
Landsat-ETM image date 2001/02/03
By the same way, cloud part in the image date 2002/01/05 was
removed based on the image date 2001/02/03. Result is shown
in the following figure: