Proceedings PaperParallax error correction techniques by image matching for ASTER/SWIR band-to-band registration
|Format||Member Price||Non-Member Price|
ASTER/SWIR is an imager to cover the 1.6 to 2.5 micrometer wavelength region. The spatial resolution is 30 m and the spectral region is divided into 6 bands. The detectors are linear Pt- Si Schottky-barrier diode CCD arrays. In order to cool focal plane assembly down to 80 k with a single cryogenic cooler, all six bands CCD arrays are allocated in parallel on a focal plane, which causes band to band misregistration as a parallax effect of ground level variation. This paper deals with an approach to evaluate the parallax value from the images by means of band to band image position matching and reduce the effect to achieve band to band registration. The approach first finds corresponding points for a pair of images among the six spectral-band images, then interprets the data into an attitude of line of sight and geometric distortion by parallax. By using these data, image distortion can be corrected to eliminate parallax effect of all six bands by a resampling process named FFT-nearest neighbor selection. By applying this method on existing images taken by actual SWIR imager, misregistration for all six bands is reduced to about 1/8 of pixel spacing as root mean square value.