Project: Fully Automated Registration and Composite
Generation of Multisensor and Multidate
Satellite Image Data
K. Mitra and
Deng ,David Garza and Leila Fonseca
M. Lee (JPL)
Ongoing research projects include multiscale registration
using robust image segmentation and matching algorithms with the
aim of developing faster registration algorithms. In addition,
efficient data representation schemes are being investigated by
integrating segmentation, registration, fusion and data
decorrelation in a single framework. Such a compact
representation would allow fast and convenient browsing and
retrievals of image data. This is an interesting problem
particularly in the context of large image databases such as the
UCSB Alexandria Digital Library project.
Contour Based Approach to Multisensor Image Registration:
Compiled registration code for NT,
SGI, SUN, IBM/PC and UNIX.
images are also provided
Explorer user interface
Multispectral (Landsat TM) Image Registration
TM - Spot) Image Registration
Research on image registration addresses issues
related to multiscale registration, integrating robust
segmentation tools with the contour matching approach developed
at UCSB, and in developing novel image representation schemes for
maintaining large databases of satellite images.
The initial focus of the project has been on the
implementation of the basic countour matching, the elastic
matching apporach, and the evaluation of the prototyped methods
using a graphical image processing user interface tool, namely
the EXPLORER image processing environment on a Silicon Graphics
workstation. The research work focused on extending the method of
Li, Manjunath, and Mitra ("A Contour Based Approach to
Multi-sensor Image Registration"). The method has been
implemented using C programming language and ported to the
EXPLORER interface. Preliminary tests of this method compared
with some standard methods have been carried out at UCSB. This
approach extracts contour information from each of two images,
correlates salient features of the contours, and finds optimal transform parameters for aligning the images. After the two images
are registered, they are combined to present the result.
Preliminary results of the fusion using the wavelet transform
have been explored and showed to be very promising. This method
works by assuming that the high frequency wavelet coefficients
have greater energy in the image that contains useful features in
that region than in the image that does not. The images are fused
by constructing a wavelet pyramid using the the more dominant
high frequency wavelet coefficients among the two images. The
fused image is supplied from the inverse wavelet transform of
this combined pyramid. This method has been ported to the
EXPLORER interface on the SGI machine. Details are described in
Li, Manjunath, and Mitra ("Multisensor Image Fusion Using
the Wavelet Transform").
The developed registration and fusion methods have been
evaluated by our collaborators in the Geography Department at
UCSB. Extensive testing of the registration algorithm on a wide
range of data has been carried out.
These materials are
presented to ensure timely dissemination of scholarly and technical
work. Copyright and all rights therein are retained by authors or by
other copyright holders. All persons copying this information are
expected to adhere to the terms and constraints invoked by each
authors copyright. In most cases, these works may not be reposted
without the explicit permission of the copyright holder.
Hui Li, B. S. Manjunath and Sanjit K.
Mitra "A contour-based approach to multisensor image registration",
IEEE Transactions on Image Processing, vol.4, (no.3), pp.320-34, March 1995.
L. M. G. Fonseca and B. S.
Techniques for Multisensor Remotely Sensed Imagery
of Photogrammetry Engineering and Remote Sensing, Sep 1996.[abstract]
H. Li, B. S. Manjunath and S. K. Mitra,
"Multisensor Image Fusion Using the Wavelet Transform,"
Proc. first international conference on image processing, ICIP 94, Austin,
Texas, Vol. I, pp. 51-55, Nov 1994.
Demo presentation (NASA workshop) 5.8Mb