Detection and texture synthesis

· 2 min read
previous-research-theme

joint work with S. Thivin (PhD 2012–2015) and M. Prenat

Detecting small, sub-resolved targets in optronic images is difficult when the background is a complex texture such as clouds. This theme corresponds to S. Thivin’s CIFRE PhD with Thales Optronics, carried out with M. Prenat.

Local stationarity tests for target detection

With S. Thivin and M. Prenat, we detected targets as points that invalidate a local stationarity assumption on the background texture. The tests rely on simple local (Markovian) models estimated on adapted domains, obtained by segmenting the image into stationary zones, and we paid particular attention to their calibration (SFdS 2014 and 2015, OPTRO 2016).

Texture synthesis for calibration

We also studied how texture synthesis algorithms can be used to better calibrate these tests. This led to a study of cloud texture synthesis and to an image synthesis patent with M. Prenat and G. Berginc (granted in 2023).

Publications

2023 (1)
Patents
  • Procédé de synthèse d'images. M. Prenat, E. Le Pennec, G. Berginc. FR3075438B1, 2023. PDF
    BibTeX
    @patent{Prenat19,
      author = {Prenat, M. and Le Pennec, E. and Berginc, G.},
      title = {Procédé de synthèse d'images},
      number = {FR3075438B1},
      year = {2023}
    }
2015 (1)
National conference proceedings
  • Détection automatique de cibles sous-résolues. S. Thivin, E. Le Pennec, M. Prenat. Journée de la SFdS, Lille, 2015. PDF
    BibTeX
    @inproceedings{ThivSfdS15,
      author = {Thivin, S. and Le Pennec, E. and Prenat, M.},
      title = {Détection automatique de cibles sous-résolues},
      booktitle = {Journée de la SFdS},
      address = {Lille},
      year = {2015}
    }
2014 (1)
National conference proceedings

Talks

2016 (1)