ESTIMATION OF ATMOSPHERIC OPTICAL
REMOTELY SANSED IMAGE CORRECTION
During a long time the possibility to estimate the
atmosphere optical parameters using remotely sensed images only has been under
studying in our Institute. The attempts have been made to estimate such
atmospheric parameters as the atmospheric optical depth τ (or the direct
transmittance Tp=exp(-τ)) and the Normalized Atmospheric Point
Spread Function (NAPSF).
ESTIMATION OF THE ATMOSPHERIC
Our method of estimating the atmospheric optical depth
(the direct transmittance) is based on the comparison between the values of
sampling variances for radiances of all pixels on two images for the same area
received at two different angles of observation .
From analyzing the error of this estimation method we
can see that in the case of a uniform field of Lambertian reflectance the
estimate agrees with the true value of the atmospheric optical depth. The deviation
of the surface from the Lambertian one gives rise to the error in estimation of
optical depth. Since the angle scattering properties of real surfaces on images
are unknown and a quality of the estimates is not controlled, we have to draw
the conclusion that the estimate of the atmospheric optical depth (the direct
transmittance) using two images acquired at two angles of observation is not
suitable in practice.
The situation becomes more preferable if we have images
of the same site acquired for several angles of observation as in the case of
the CHRIS spectrometer for example. We have managed to develop the controllable
procedure of the atmospheric optical depth (direct transmittance) estimation
using images for several angles of observation .
In Fig.1 you can see the direct transmittance estimates
obtained by us for the first twelve spectral channels of the CHRIS sensor.*
Here we also depicted the average values of the atmospheric direct transmittances
for the same twelve spectral channels of the CHRIS as the results of simulation
by means of the LOWTRAN model for four assumed aerosol conditions. Comparing
the estimates obtained with the simulation results we can see their similarity
on the whole. We think that this preliminary result is hopeful and we are ready
to collaborate with the researcher who is interested in this subject.
1. Direct transmittance estimation results for 12 channels of the spectrometer
ESTIMATION OF THE NAPSF
It’s known that the atmosphere can be considered as a
passive linear low-pass spatial filter that smoothes the images of the Earth
(so-called “adjacency effect”). Such filter-atmosphere is completely determined
by its Atmospheric Point Spread Function (APSF). We have managed to develop a
direct method of estimating the discrete Normalized Atmospheric Point Spread
Function (NAPSF) from available images alone, not using any external sources of
information about the atmosphere. Knowledge of the NAPSF enables us to synthesize
the correction filter, which is able to remove the adjacency effect from
remotely sensed data.
Detailed description of the NAPSF estimation method and
adjacency effect correction suggested by us can be found in [3-4].
Figure 2 shows the result of adjacency effect correction for a
fragment (512x512) of a real image acquired in 2nd spectral channel
(0.546μm – 0.556μm) of the spaceborne CHRIS sensor. The original
fragment is shown in the left picture, while the result of image correction is
depicted in the right frame.
Figure 3 shows a fragment (512x512) of the original
image acquired in sixth spectral channel (0.54μm) of the airborne
hyperspectral sensor MIVIS (left picture) and the outcome of the correction
filtering (restored image) (right picture).*
We are open to any form of collaboration with
specialists on the atmospheric point spread function estimation problem and
adjacency effect correction.
A.A. Semenov, “Estimation of atmospheric
correction parameters in analyzing of multispectral scanner images”, The
Earth research from space, no.2, pp.38 – 45, (2002). (Russian edition).
A.V. Moshkov, V.N. Pozhidaev, and A.A. Semenov,
“Estimation of the transparency from the Earth’s surface images obtained in
optical range at different angles”, Journal of communications technology and
electronics (English translation of Radiotekhnika i electronika), v. 54,
no. 9, pp.1000 – 1002 (2009).
A.A. Semenov, “Correction of the scanner image
distortions due to the adjacenc effect”, The Earth research from space,
1, pp.35 – 48 (2004) (Russian edition).
A.A. Semenov, A.V. Moshkov, V.N. Pozhidaev, A.
Barducci, P. Marcoionni, and I. Pippi, “Estimation of the normalized
atmospheric point spread function and restoration of remotely sensed images”, IEEE
Trans. Geosci. Rem. Sens., vol.49, no.7,pp. 2623 – 2634 (2011).
* Applicable CHRIS and MIVIS images
were kindly presented by our colleagues from the Instituto di Fisica Applicata
“Nello Carrara”, Florence, Italy.
You are welcome to
connect with us.
Alexander Semenov: E-mail:
Victor Pozhidayev: E-mail: