3D Texturing From Multi-Date Satellite Images

Enric Meinhardt-Llopis, Marie d'Autume; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2019, pp. 0-0

Abstract


This paper addresses the problem of point cloud textu-ration of urban areas from multiple satellite images. Ouralgorithm is well-suited for the case where the images areobtained from different dates, where the dynamic ranges areincompatible and the position of the shadows is different.The method relies on a robust, PDE-based, fusion of themultiple candidate textures for each surface patch, and op-tionally on a geometric criterion to remove the shadows ofthe known objects. We showcase the results by building a3D model of some areas of Boulogne Sur Mer (Argentine) such that all facades are correctly textured, with uniformcolours and without shadows, even if for each individualinput image only one side of the buildings was visible.

Related Material


[pdf]
[bibtex]
@InProceedings{Meinhardt-Llopis_2019_ICCV,
author = {Meinhardt-Llopis, Enric and d'Autume, Marie},
title = {3D Texturing From Multi-Date Satellite Images},
booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops},
month = {Oct},
year = {2019}
}