iBHSIHHMHnHRSnHBMKfc*
COMPRESSION SPECIFICATION FOR EFFICIENT USE OF HIGH RESOLUTION
SATELLITE DATA
Emmanuel Christophe, Carole Thiebaut and Christophe Latry
CNES DCT/SI/AP Bpi 1219
18 av E. Belin, 31401 Toulouse Cedex 09
{emmanuel.Christophe, carole.thiebaut, Christophe.latry}@cnes.fr
Commission IV/9
KEY WORDS: Compression, high resolution,optical, JPEG 2000, data transfer
ABSTRACT:
For an efficient usage and distribution of high resolution satellite images, several problems need to be solved. The issue comes with
the increasing size of these data. Constraints are different than those of the on-board compression, thus different solutions can be
selected. For on-board compression, the main constraints are the computational complexity and the rate attainable with qualified space
equipments. For on-the-ground compression, computational constraints are not so strong, but particular care is needed to make sure
that the chosen format is widely spread and that users will be able to exploit these data easily.
1 INTRODUCTION
3 END-USERS REQUIREMENTS
For an efficient usage and distribution of high resolution satel
lite images, several problems need to be solved. The issue comes
with the increasing size of these data. A standard scene from
the Pleiades satellite, the coming optical high resolution satel
lite from CNES, will typically represent 14 GB of data (more
4 x 4 pixels in 4 spectral bands on 16 bits). The difficul
ties with such data do not occur only in the distribution process,
but also in the processing steps, data selections, quality assess
ment.
End-users need the access to derived products from the satellite
image. For a question of cost or efficiency, the high resolution
is often not required on the full scene. One of the most valuable
product is the high resolution on a specific area (city, disaster
area,...) and a lower resolution around to give a precise idea of the
context of the scene. Ideally, this higher resolution information
should be available on-demand according to the user’s decision
on the lower resolution level.
Constraints are different than those of the on-board compression,
thus different solutions can be selected. For on-board compres
sion, the main constraints are the computational complexity and
the rate attainable with qualified space equipments. For on-the-
ground compression, computational constraints are not so strong,
but particular care is needed to make sure that the chosen format
is widely spread and that users will be able to exploit these data
easily.
Fast and multiresolution data access is also necessary. When vi
sualized at full resolution, one scene represents more that 700
computer screens. In these conditions, finding the valuable infor
mation is like looking for a needle in a haystack.
This multiresolution feature is also very valuable for the steps of
data quality assessment. For example, the full resolution is not
necessary when establishing the cloud coverage notation.
2 SATELLITE DATA SPECIFICITIES
Most of popular compression formats cannot be used for satellite
data which often reach their limits. The first problem is the size
of the data: already often reaching 30000 pixels per dimension,
it is easy to foresee products exceeding 2 1 pixels. The second
problem is the dynamic range of the values: instead of the more
common 8 bits, these data are often coded on 12 bits to be able
to cover the wide dynamic range of observed scenes. Finally,
these data often comprise at least four spectral bands while most
compression formats target the typical color images with three
components (Fig. 1).
Another strong requirement is the possibility to include complex
metadata directly on the compressed streams. Apart from the
essential localization information, additional data corresponding
to some data extraction results (classification, road extraction,
clouds mask...) or coming from vectorized map should be avail
able in the images.
In some situations, fast access to the data is critical. For example,
when the image is to be used in the context of a crisis (hurricane,
tsunami, flooding, earthquake) to help to organize the rescue, it is
inconceivable to waste precious hours in inefficient data transfer.
The JPEG 2000 standard was designed to answer these limita
tions (Taubman and Marcellin, 2002). The facts that this is a well
accepted standard and that it is starting to be widely available
through numerous and well performing implementations make
it a good candidate for satellite data distribution. However, this
standard comprises many options and not all are adapted for high
resolution satellite images. Thus, it is necessary to explore these
possibilities to find the right combination ensuring an efficient
access for the end-user with a limited complexity.
New data distribution schemes can also arise from advanced data
structures. For example, the full product could be delivered on a
DVD with part of it encrypted. Unlocking the full product would
consist in buying the key from the data provider which results in
a near instantaneous transaction. The main advantage is that the
low resolution data can be delivered for a very reasonable price
and the full resolution with a very short delay according to the
image processing requirements.
1283
111
mmftp's
h%
Æ
mi
m
m
JùV.
'wmm
m
S® '
IIÖ
№
WMÊ
HHH
w
111
»r
fl
■ №{
in