Automated co-registration of images from multiple bands of Liss-4 camera

Abstract

Three multi-spectral bands of the Liss-4 camera of IRS-P6 satellite are physically separated in the focal plane in the along-track direction. The time separation of 2.1 s between the acquisition of first and last bands causes scan lines acquired by different bands to lie along different lines on the ground which are not parallel. Therefore, the raw images of multi-spectral bands need to be registered prior to any simple application like data visualization. This paper describes a method for co-registration of multiple bands of Liss-4 camera through photogrammetric means using the collinearity equations. A trajectory fit using the given ephemeris and attitude data, followed by direct georeferencing is being employed in this model. It is also augmented with a public domain DEM for the terrain dependent input to the model. Finer offsets after the application of this parametric technique are addressed by matching a small subsection of the bands (100×100 pixels) using an image-based method. Resampling is done by going back to original raw data when creating the product after refining image coordinates with the offsets. Two types of aligned products are defined in this paper and their operational flow is described. Datasets covering different types of terrain and also viewed with different geometries are studied with extensive number of points. The band-to-band registration (BBR) accuracies are reported. The algorithm described in this paper for co-registration of Liss-4 bands is an integral part of the software package Value Added Products generation System (VAPS) for operational generation of IRS-P6 data products.