A MapReduce-based distributed and scalable framework for stitching of satellite mosaic images

Satellite mosaic images are of huge sizes and, therefore, the stitching process becomes time- and resource-consuming. To overcome this challenge, we propose a MapReduce-based distributed and scalable image stitching framework. In this framework, we first convert all raster images to binary format. Each binary image is compared with other images using the Point Set Pattern Matching algorithm (PSPM). Each image pair is sent to the mappers along with their position information. Mapper nodes compute a similarity number for each position of each image pair. The reducer node finds out the best stitching position for each image pair by using corresponding similarity values provided by the mappers. The position with the highest similarity value is the best position for the stitching. When there are more than two images, a specific overlap graph is created whose nodes represent images and edges represent image pairs to be stitched. The graph is created by using the position knowledge which is obtained during the computation of the highest similarity numbers between each possible image pair. The performance of the proposed framework is tested on a variable number of different-sized input images on a cluster in terms of speedup and efficiency. © 2021, Saudi Society for Geosciences.