Cloud Detection by Luminance and Inter-band Parallax Analysis for Pushbroom Satellite Imagers
Tristan Dagobert, Rafael Grompone von Gioi, Carlo de Franchis, Jean-Michel Morel, Charles Hessel
→ BibTeX
@article{ipol.2020.271,
    title   = {{Cloud Detection by Luminance and Inter-band Parallax Analysis for Pushbroom Satellite Imagers}},
    author  = {Dagobert, Tristan and Grompone von Gioi, Rafael and de Franchis, Carlo and Morel, Jean-Michel and Hessel, Charles},
    journal = {{Image Processing On Line}},
    volume  = {10},
    pages   = {167--190},
    year    = {2020},
    doi     = {10.5201/ipol.2020.271},
}
% if your bibliography style doesn't support doi fields:
    note    = {\url{https://doi.org/10.5201/ipol.2020.271}}
published
2020-11-21
reference
Tristan Dagobert, Rafael Grompone von Gioi, Carlo de Franchis, Jean-Michel Morel, and Charles Hessel, Cloud Detection by Luminance and Inter-band Parallax Analysis for Pushbroom Satellite Imagers, Image Processing On Line, 10 (2020), pp. 167–190. https://doi.org/10.5201/ipol.2020.271

Communicated by Charles Hessel
Demo edited by Tristan Dagobert

Abstract

This paper proposes a cloud detection algorithm for Earth observation images obtained by pushbroom satellite imagers. The pushbroom technology induces an inter-band acquisition delay leading to a parallax effect for the clouds. We propose a method exploiting this characteristic thanks to the analysis of the inter-band disparity. Several other features discriminating clouds are also defined and all are merged to build a robust a contrario statistical decision. Experiments applied on scenes acquired by various pushbroom satellites such as Sentinel-2, RapidEye and WorldView-2 show the effectiveness of the proposed method. In particular, we demonstrate a balanced accuracy rate close to 98% for cloud and non cloud classification for Sentinel-2 images.

Download

Non-Reviewed Supplementary Materials

These files and information are provided by the authors and have not been reviewed.

It consists of 9854 RGB patches of size 366x366, extracted from the original tiles of size 10980x10980 with their associated ground truth.