The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B5. Beijing 2008 
■ 
int BlockRow; 
//the row number of pixels in a Block 
int BlockColumn; 
//the column number of pixels in a Block 
double BlockSize; 
//the Block size (unit: degree) 
double BlockCellSize; 
//the resolution of pixels (unit: degree) 
} 
We can find that the relationship between the parameter of i+1 
layer and i layer is as follows: 
XBlocksNum i+1 = 2*XBlocksNum i 
YBlocksNum i+1 = 2*YBlocksNumj 
BlockRow i+1 = BlockRow; 
BlockColumn i+1 = BlockColumn; 
BlockSize i+1 = BlockSize;/2 
BlockCellSize i+ i= BlockCellSize; /2 
So every tile in the pyramid has the same number of pixels 
which helps to make the system running effective. 
method is the improved method based on EZW encode method. 
It can generate an embedded bit stream, and when receiving the 
bit stream it can break the received bit stream at any time for 
reconstruction. So it has good progressive transmission 
character. 
This paper introduces an improved SPIHT encode method, it 
firstly evaluate the terrain surface complexity, and then 
calculate the bit rate of the encoding process according to the 
terrain surface complexity and terrain scale. So the terrain data 
in different complexity can be compressed effectively. 
3.1 The relationship between wavelet compression and 
terrain surface complexity in visualization 
The mathematic mechanism of DEM compression mainly 
includes the two points: 
The first point is that the information of origin data exists big 
redundancy. For example the DEM data has elevation relativity 
in the adjacent grids. The information redundancy will generate 
extra coding. If we get rid of this redundant information the 
space take up by information will be reduced [8]. 
And we can get the RowID and ColumnID of block and pixel 
using the geographic coordinate (-180° <B<180° , -90° 
<L<90° ): 
xblock = (int) 
xpixel = (int) 
ypixel = (int) 
2?+ 180 f 
BlockSize i 
5 + 180 
yblock = (int)- 
1 + 90 
BlockCellSize i 
L + 90 
BlockSize i 
- xblock x BlockRow. 
BlockCellSize i 
- yblock x BlockColumn i 
Once the global quad tree is built, each cell in a layer 
corresponds to a certain longitude and latitude degree. If we 
need to add new data into it, we only need to update the node of 
quad tree; the structure of quad tree doesn’t need to be modified. 
We can conclude that the tree structure has superiority in such 
aspects: 
1) Data redundancy is greatly avoided 
2) The visualization is simplified and the computer resource 
is saved. 
3) All the data can be partitioned according to the longitude 
and latitude. So it can be easily accessed 
4) It is good in expanding; you can add the higher resolution 
data as you want without the quad tree structure altered. 
3. DEM COMPRESSION METHOD 
The second point is that DEM in very high precision is not 
necessary in some application area. For example in DEM 
visualization human eye is the information receiver, it cannot 
perceive the tiny hypsography. So in high compression ratio the 
decompressed DEM data still obtains satisfactory subjective 
quality. 
Wavelet method is effective in removing data spatial relativity. 
After wavelet transformation the data amount is the same. But 
the information energy is reallocated. Above 95% of energy 
centralize in the low frequency part; it describes the rough 
sketch of the terrain surface. Other high frequency parts 
describe the detailed component of the terrain surface. The 
principle of wavelet compression is to adopt approximate 
coefficient in low sampling rate and some approximate 
coefficient which we are interested to approach the origin 
terrain data. If we obtain more approximate coefficients, the 
distortion of decompressed data is low, but the compression 
ratio is low; if we obtain less approximate coefficients, the 
compression ratio is high, but the distortion of decompressed 
data is high. How to balance the compression ratio and the 
decompressed data quality is what we should research. 
In the research we can find that when the DEM region is flat, 
the energy of the detailed component of the terrain surface is 
low, and we can adopt higher compression radio in the wavelet 
coding method with relative small distortion; and when the 
terrain surface is mountainous and fragmentized, if we adopt 
the same compression ratio as the flat region, the distortion will 
be huge. 
/1 
The DEM data in global region has huge data amount, which 
brings great challenge to DEM storage, transmission and real 
time rendering. Therefore how to compress and simplify DEM 
data is one of the key techniques. 
At present the DEM compression method mainly includes 
converting grid structure to tin structure, entropy encode 
method, and mature image compression method [7]. 
Nowadays the wavelet transformation has been successfully 
adapted in the video and image compression. SPIHT encode