Relative radiometric normalization for mosaicking IRS CartoSat-2 panchromatic images using genetic algorithm

Abstract

Urban features can be analysed using high-resolution satellite images. Satellites procure panchromatic images at a spatial resolution even better than 1 m. However, it cannot cover the large area of a city in one scene. Therefore, satellite images from different dates and times must be mosaicked to study the urban region as a whole. Efficient and seamless mosaicking requires relative radiometric normalization (RRN), which generally is an iterative process. In this paper, an approach using the genetic algorithm has been developed for normalization in a single step. The algorithm has been tested on Jaipur city in India. Results produced for the mosaic using genetic algorithm provide visually seamless mosaic over the study region with significant improvement of the RMSE among the images’ overlap region. The algorithm has been compared against the mean-standard deviation (MS) and histogram matching (HM) methods of RRN with average RMSE improvement of 40% and 38%, respectively.