Automated image mosaicking

Abstract

This paper developed and implemented automated image mosaicking procedures to produce seamless and smooth mosaics from random sequences of digital aerial images and satellite images. The procedure employs automated algorithms for global image registration, seam line extraction, radiometric adjustment, and composition of overlap images. Results of the experimentation on mosaicking black and white orthorectified aerial images and on mosaicking multi‐resolution color infrared SPOT‐2 and Landsat‐5 satellite images show that the implementation of automated conjugate point extraction and image matching algorithms automates the establishment of geometric registration among overlapped images, while the seam line selection and two‐step lookup table histogram matching approaches facilitate creating globally smooth and qualitative mosaics. The system developed in this paper makes complex composition of multi‐source multi‐resolution multi‐band images with different geometric and radiometric characteristics an easy and efficient solution for producing reliable mosaic images.