Appication of Wavelet Transform for cloud Removal On Spot 2/4 Image

No image available for this title
This thesis explore application of wavelet transform for cloud removal on SPOT 2/4
data. Wavelet transform is a tool for data convertion from time domain to frequency domain.
Wavelet transform is different with Fourier transform that time information is still remain.
The application of wavelet transform implement a technique of wavelet decomposition which
separate low frequency and high frequency components. Since there is time information, in
the application for image, pixel based operation could be performed.
Cloud removal use the advantage of wavelet decomposition. Wavelet decomposition is
used to extract feature of cloud. Cloud is supposed to have smooth and flat intensity over their
body, and shadow either. Here, there is no high frequency information. Thus using wavelet
decomposition, cloud or shadow could be detected in pixels which have low values for high
frequency components.
Previous research on this topic is not dealing with haze or the edge of cloud. High
frequency components exist in that area, then the concept of low frequency could not be used.
At the edge, intensity deviate sharply, then high frequency components have very high value.
This characteristic is used for improvement to previous research.
There are several option to perform the technique of cloud removal. Selection of wavelet
bases is important. db4 is used since approximation and wavelet coefficient are more smooth
than the simplest wavelet bases, Haar wavelet. The use of 2D-Stationary Wavelet Transform
instead of 2D-Discrete Wavelet Transform is in purpose to get more smooth and accurate
result since there is no downsampling. Level 1 of decomposition is satisfactory though using
higher level of decomposition may provide better result.
The method of this research is compared with conventional method. Conventional
method apply threshold to intensity for cloud, as well as for shadow. Two images are used to
test the methods. Using SPOT 2/4 data, the difficulty comes when the image is not
orthorectified and not brightness corrected. Since grayscale is used, in some places, land
surface has high value of intensity which is similar with clouds. The same case happen to
shadow detection. The proposed method using wavelet doesn’t have good result for shadow
detection. For cloud detection, more cloud pixels could bedetected, but false detection is also
higher. Then combination of wavelet method for cloud detection and conventional method for
shadow detection is performed. Both combined method and conventional method have their
own advantage. For combined method, cloud detection using wavelet have easiness to select
threshold value, but selection of threshold value in conventional method depends on
illumination. Conventional method has the advantage that error detection is lower. In the
future, further research on wavelet may reduce error detection in combined method and
furthermore better wavelet-based algorithm might be found for shadow detection.
Key Words : Wavelet transform, cloud removal, SPOT, wavelet bases, wavelet
decomposition, image fusion, cloud detection, shadow detection
Informasi Detil
Judul Seri -
No. Panggil 01/2017/THE
Penerbit BUAA : Beijing.,
Deskripsi Fisik 59P., Ill., Pdf.
Bahasa English
ISBN/ISSN 10006LS0825316
Klasifikasi TP391
Edisi -
Subyek SPOT
Wavelet Bases
Wavelet decomposition
Wavelet Transform
Cloud Removal
Cloud detection
Image fusion
Pernyataan Tanggungjawab
Beihang University, Beijing, China
Tidak tersedia versi lain

Pilih Bahasa

Tipe Koleksi Dipublikasikan

Tipe Koleksi Tidak Dipublikasikan

Advanced Search


This software and this template are released Under GNU GPL License Version 3.