Global seamline networks for orthomosaic generation via local search

A novel method for establishing a seamline network for orthomosaic generation is presented. The seamline network determines the regions of the mosaic allocated to each of the original images. Our method allows for optimal network vertices to be selected while avoiding a combinatorial optimisation task. The computational requirements (time and storage) of our proposed method are linear in the number of images. Seamlines between vertices are found using a graph-based approach which finds shortest paths, subject to a constraint which minimises the maximum cost of any single edge. Experiments on synthetic and real-world data sets illustrate the value of the approach compared to naïve vertex selection and recently proposed methods.