It is well recognized within the remote sensing community that there exists spectral correlation between bands 1, 2, and 3 (which represent reflected blue, green, and red light respectively), and also bands 5 and 7 (which are reflected middle-infrared bands) of the Landsat Thematic Mapper (TM) multispectral image. In this paper, we are presenting the outcome of some experiments which use the spectral correlation as well as spatial correlation of the brightness values (BVs) to compress the TM multispectral data. Our method compresses one of the bands using the standard JPEG compression, and then orders the next band's data with respect to the previous band's sorting permutation. Then, a move to front coding technique is used to lower the source entropy, before actually encoding the data. It has been observed that our method yields tremendous gain on the visible bands (on the average 0.4-0.6 bpp) and can be successfully used for multispectral images where the spectral distances are closer.
[1]
Khalid Sayood,et al.
Data compression in remote sensing applications
,
1992
.
[2]
P. W. Jones,et al.
Digital Image Compression Techniques
,
1991
.
[3]
Gregory K. Wallace,et al.
The JPEG still picture compression standard
,
1991,
CACM.
[4]
Arun N. Netravali,et al.
Digital Pictures: Representation and Compression
,
1988
.
[5]
Jindi Wang,et al.
Spectral and spatial decorrelation of Landsat-TM data for lossless compression
,
1995,
IEEE Trans. Geosci. Remote. Sens..
[6]
Glen G. Langdon,et al.
Universal modeling and coding
,
1981,
IEEE Trans. Inf. Theory.
[7]
Stephen R. Tate,et al.
Band ordering in lossless compression of multispectral images
,
1997,
Proceedings of IEEE Data Compression Conference (DCC'94).
[8]
John R. Jensen,et al.
Introductory Digital Image Processing: A Remote Sensing Perspective
,
1986
.