International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B7. Istanbul 2004 
4. CONCLUSIONS AND DISCUSSIONS 
The 3D wavelet is a multi-band wavelet analyse and it can 
obtain edge and structure feature with higher accuracy, which is 
very use for image compression to obtain high reconstruct 
image quality. In this paper, the 3D wavelet transformation is 
proposed for the edge reservation image compression. The 
experiments of three typical multispectrum sensor obtain over 
than 16 compression ratio. The reconstruct image quality can 
meet the quassi-lossless requirement. It can be use for the image 
database buiding and image transmitting. The further 
improvements of the proposed method can be summarized as 
followings: 
(1) Edge extraction algorithm: to obtain high precision edge 
and structure feature of the objects in the multispectrum image 
(2) Quantification poly to wavelet coefficient 
ACKNOWLEDGEMENT 
The author would like to thank MODIS Research Center to 
provide MODIS images. 
REFERENCES 
[2] A. K. Jain, Fundamentals of digital image processing, 
Prentice-Hall International Editions, 1991. 
[3] J. A. Saghri, et:al:, Practical Transform Coding of 
Multispectral Imagery," IEEE Signal Processing Magazine, pp. 
32-43 Nov. 1995. 
ZHOU S T, XUAN J B. Restoration Technique Based on Image 
Features. The Journal of Wuhan Technical University of 
Surveying and Mapping, 1999.V01.24(3):230-234. 
WU L N. Principle and Application of Data Compression. 
Beijing: Publishing House of Electronic Industry, 1994. 
ZHU X CH, HU D. Digital Image Communication. Beijing: 
Publishing House of the People’s Posts and 
Telecommunications, 1995. 
HU Qingwu, Technique of Quasi-lossless Compression of 
Multiple Spectrum Remote Sensing Images Based On Image 
Restoration. SPIE: 4551-37,2001 
Zhang Rong , Yan Qing , Liu Zhengkai . A Prediction Tree 
based Lossless Compression Technique of Multispectral Image 
Data [ J ] .Journal of Remote Sensing , 1998 , 2 (3) : 171-175. 
Zhang Rong , Liu Zhengkai , Li Hougiang. Classification based 
lossless compression of multispectral image [J ]. Journal of 
Image and Graphics , 1998 , 3(2) : 106 110. 
Hoffman R N , Johnson D W. Application of EOF' s to 
multispectral imagery: data comp ression and no ise detection 
for AV IR IS [J ]. IEEE Trans Geosci Remote Sensing, 1994, 
32 (1): 25 34 
Memon N D, Sayood K, N agliras S S. Lossless compression of 
multispectral image data [ J ]. IEEE Trans Geosci Remote 
Sensing, 1994, 32 (2) : 282 — 289 
W ang J F, Zhang K, Tang S. Spectral and spatial decorrelation 
of LandsatTM data fo r lo ssless compression [J ]. IEEE Trans 
Geosci Remo te Sensing,1995, 33 (5) : 1277— 1285 
Rao A K, Bhargava S. M ultispectral data comp ression using 
60 
bidirectional interband prediction[J ]. 
Remote Sensing, 1996, 34 (2) 
Said A Pearlman W A. A n image multiresolution 
representation for lossless and lossy comp ression [ J ]. IEEE 
Trans Image Processing, 1996, 5 (9) : 1303— 1310 
IEEE Trans Geosei 
WU Jianhua, Image data compression of GSM satellite image 
[J ]. Journal of Image and Graphics, 1999, 4 (1) : 56— 60 
Dep 
KEY WC 
ABSTRA 
New sens 
tremendo 
dimensio: 
effective 
It has bee 
Dimensic 
based tec 
Extractio 
computat 
usually n: 
Spectral 
signature 
Transfort 
PCA, foi 
accuracy 
Multispe 
surface : 
collect s 
sensor t 
been im 
bands, 
example 
provide 
Theoret 
abilities 
classific 
multispe 
hypersp 
As the : 
the nun 
for ima 
proved 
related 
square 
estimat 
training 
have ar 
to perf 
are insi 
case of 
inaccur