2-35
pan-sharpening algorithms like UNB etc. Other standard
quickbird imagery products are used to research and develop
the new pan-sharpening algorithm too.
3. METHODOLOGY
This can be explained and described by the point-spread
function (PSF). Under a slant viewing angle, the cross
radiation of the adjacent pixels side to the spacecraft are more
obvious than the other side (XU, 2005). On the other side, the
more short of the wavelength, the more apparent is the cross
radiation.
3.1 Atmospheric effect and Quickbird sensor ’ s System
MTF
Vegetation has the strongest reflection at the near-infrared
range and considerable absorption at the visible bands. On the
contrary, water, roads and other objects like house roof etc
absorbs the near-infrared radiation, and reflects most of the
visible radiation. The great differences in the near infrared and
visible reflection properties of geographic features make the
coupling between the features and other object around the
destination. In this process, atmosphere plays a critical role.
Atmospheric cross radiation effect can be described by the
point spread function (PSF), whereas by the MTF in the
frequency domain. The Quickbird sensor’s System MTF
includes atmosphere, telescope, detector, and data
compression/expansion (Paul W.Scott, 2004). The Diffraction
from optics also has contribution to MTF. Thereupon, the GSD
of the sensor at nadir in fact is smaller than the nominal
resolution.
3.2 PSF and Misregistration between pan and ms bands
Through constructing a new PSF according to the geometry
relation between the image scene and the sensor, we can solve
this problem. The analytic expression of the PSF is very
complex and not convenient to use in fact. We constructed the
new PSF with a form
of 5*
5 matrix
shown as follows.
0
0
0.05
0
0
0
0.12
0.47
0.30
0
PSF =
0.02
0.42
7.00
0.60
0.08
0
0.12
0.47
0.30
0
0
0
0.05
0
0
Where sensor’s azimuth angle = 90
The new PSF is convenient to use in computing with some RS
software like EADARS etc. The matrix can be rotated with the
sensor’s azimuth angle.
Misregistration between pan and ms bands results in a blurry
pan sharpened image. There are many factors can cause
misregistration. These factors include random spacecraft
motion between collects, parallax between the bands and
pan-sharpening Algorithm like PCA (Figure 2) and Wavelet
sharpening.
3.3 The establishment of the new pan-sharpening
algorithm
A. Pre-processing of Quickbird data
All pan and ms bands of the image scene will be used in this
study. The DNs have been converted into radiance value. The
conversion equation (Kei Krause, 2003) is as follows.
L = absCalFactor □Q (2)
pixel .band band ^pixel .band v '
Where w =top-of-atmosphere spectral band
-integrated radiance image pixels
absCalFactor■ =absolute radiometric calibration
factor for a given band and is listed in the .IMD file
Q w =radiometrically corrected image pixels
(DNs)
Figure 2. Pan-sharpened image with PCA algorithm
(misregistration is apparent around bright objects)
(satAz = 298.4;satEl =69.6;crossTrackView Angle =
-18.1;inTrackViewAngle = 5.9;)
In many pan-sharpened image (Figure 4) with PCA algorithm,
the blurriness around bright objects always orientates to the
sensor’s observation angle , so we can simply get the sensor’s
(not satellite’s) azimuth angle with respect to the scene from
the pan-sharpened image. Thereupon, the sensor’s orientation
to the scene is the main factor that causes a misregistration
between the pan and MS bands and result in a blurry that is
apparent around bright objects in fusion process.
There are many resampling method to upsample the MS bands
GSD to the PAN bands. Cubic Convolution can slightly
increase or decrease MTF of the image data (Paul W.Scoott,
2004). So in the new pan-sharpening algorithm, we select the
Cubic Convolution as the main resampling method.
B. Building of the new pan-sharpening algorithm
Since the panchromatic band covers all ms bands, we can get a
approximated panchromatic band from the all four ms bands
based on the QB spectral band responses (Figure 1) and the
Top-of-atmosphere sun radiance in the QB spectral range. This
can be shown as an equation as follows.