How are the enhanced resolution image products produced?
Resolution enhancement is done using the BYU-developed Scatterometer
Image Reconstruction (SIR) with Filtering (SIRF) algorithm. The
SIRF algorithm was originally developed to enhance Seasat scatterometer
image resolution by combining data from multiple passes of the satellite
(Long, Hardin and Whiting, 1993) but has also be used with SSM/I
radiometer data (Long and Daum, 1997) and ERS scatterometer data
(Long, et. al, 1994). A number of improvements to the original SIRF
algorithm have been developed to optimize its performance.
The SIRF algorithm is based on a multivariate form of block multiplicative
algebraic reconstruction. Combining multiple overlapping passes
and robust performance in the presence of noise, it provides enhanced
resolution measurements of the surface characteristics. The method
used is a true reconstruction of the surface response using
information in the sidelobes of the measurement resonse function
(Early and Long, 2001).
To provide a simple intuitive explanation of the idea behind SIRF,
consider the following. (The incidence angle dependence of sigma-0
is ignored in the following discussion.)
Let f(x,y) be a function that gives the surface sigma-0 at a point
(x,y). The scatterometer measurement system can be modeled by
z = H f + noise
where H is an operator that models the measurement system (sample
spacing and aperture filtering) and z represents the measurements
made by the instrument sensor. The set of measurements z are a discrete
sampling of the function f convolved with the aperture function
(which may be different for each measurement). A particular measurement
z_i can be written as
z_i = Integral h_i(x,y) dx dy + noise
where h_i(x,y) is the measurement response function (due, for
example, to the antenna pattern and the Doppler filter response)
of the i-th measurement. For resolution enhancement, we are interested
in the inverse problem:
f_estimate = Inverse(H_estimate) z
where f_estimate is an estimate of f from the measurements z.
The inverse of the operator H is exact only if H is invertible and
the measurements are noise free; otherwise, the result is an approximation
to the original surface.
This represents a form of resolution enhancement since information
in the sidelobes of the measurement response or aperture function
is recovered in the inversion. In effect, this is what iterative
SIRF algorithm does, producing images at a finer resolution than
the original measurements. Thus SIR is a true resolution enhancement
algorithm which extracts information from the sidelobes of the measurement
response function to generate the final image product (Early and
Long, 1999)}; in effect, it is an inverse reconstruction filter
optimized to minimize noise in the reconstructed image.
References:
- D.G.Long, P.Hardin, and P.Whiting, "Resolution Enhancement
of Spaceborne Scatterometer Data," IEEE Trans. Geosci. Remote
Sens., vol. 31, pp. 700-715, 1993.
- D.G.Long and D.Daum, "Spatial Resolution Enhancement of SSM/I
Data," IEEE Trans. Geosci. Rem. Sens., vol. 36, pp.
407-417, 1997.
- D.G.Long, D.Early, and M.R.Drinkwater, "Enhanced Resolution
ERS-1 Scatterometer Imaging of Southern Hemisphere Polar Ice,
Proc. Int. Geosci. Rem. Sens. Sym., Pasadena, California,
8-12 August, pp. 156-158, 1994
- D.G.Long and M.R.Drinkwater, "Cryosphere Applications of
NSCAT Data," IEEE Trans. Geosci. Remote Sens., Vol.
37, No. 3, pp. 1671-1684, 1999.
- D.S.Early and D.G.Long,"Image Reconstruction and Enhanced
Resolution Imaging From Irregular Samples," IEEE Trans.
Geosci. Remote Sens., Vol. 39, No.2, pp. 291-302, Feb.
2001.