COMPARISON OF CONTOUR FEATURE BASED AND INTENSITY BASED INSAT-3D MET IMAGES COREGISTRATION FOR SUB PIXEL ACCURACIES

Image registration in meteorological images that are acquired continuously for their use in weather forecast activities and other related scientific analysis is a critical requirement. Meteorological images are obtained from geostationary orbits in visible, infrared, water vapor channels covering a large frame of several hundreds of kilometres of geographical extent which generally involve bidirectional scanning to cover larger extents. The acquired images have to be guaranteed for their geometric fidelity to a standard of choice among themselves by image registration. Registration of such images require to deal with low contrast, cloud and snow occlusions apart from navigation data uncertainties. Nevertheless, sub pixel accuracies are demanded for image analysis and geophysical parameters derivations. Feature based registration techniques are commonly used and intensity based techniques are also put to use in these contexts rarely. The proposed feature based approach uses a land water boundary data extraction with phase correlation of image blocks and proposed the intensity based approach tackles the same problem without any preprocessing step using a sampler-metric-transform-optimizer procedure. A comparison of these two approaches is pursued here in this article using various channel data sets of INSAT-3D satellite for sub pixel accuracies.