Résumé
La turbulence atmosphérique perturbe ľobservation à haute résolution. C’est un phénomène étudié depuis longtemps, en astronomie notamment. Le présent article porte sur le cas de ľobservation ďun objet situé environ vingt kilomètres, la propagation étant horizontale et près du sol, en infrarouge. Les images à longue pose sont restaurées avec des algorithmes classiques de déconvolution. Les résultats ne sont satisfaisants que pour une faible perturbation. Il est plus avantageux ďexploiter des images courte pose, car elles contiennent plus de hautes fréquences spatiales; mais ľobjet observé y fluctue aléatoirement. On travaille donc ici partir ďune séquence de plusieurs dizaines ďimages. Deux approches sont proposées. L'une consiste à analyser statistiquement les régions de chaque image pour en extraire les plus représentatives et reconstituer ainsi ľobjet réel. Les résultats sont alors très intéressants. La seconde est une approche frontière. Une analyse du problème, basée sur la méthode du maximum de vraisemblance appliquée aux descripteurs de Fourier de la forme, permet ďétablir que le contour le plus probable est obtenu par la moyenne de la position de chaque point de contour au long de la séquence. Nous déterminons ce contour en appliquant un contour actif statistique de type région, appelé « snake », optimal au sens de la théorie statistique, que nous faisons évoluer avec ľobjet au cours de la séquence.
Abstract
Atmospheric turbulence perturbs high-resolution observation. This phenomenon has been studied since a long time, particularly in astronomy. We work here on the observation of an object lying about twenty kilometer far from an infrared imaging system; the light propagates horizontally near the ground. Long-exposure images are restored by classical deconvolution algorithms. Results are satisfying only for a weak perturbation. It is better to work on short-exposure images since they retain more high-frequencies; but the observed object fluctuates randomly there, so we work on a sequence of a few tens images. Two approaches are proposed. The first one consists in a statistical analysis of each image region so as to grab the most representative of them and thus rebuild the real object. The results are here clearly better. The second one is based on frontiers. A maximum-likelihood based analysis of the Fourier descriptors of the shape shows that the optimal frontier is given by the mean position of each object edge point through the sequence. It is obtained with a region-based active contour (optimal snake, in a statistical sense) that evolves with the object through the sequence.
Références
Kolmogorov (A.N.), The local structure of turbulence in incompressible viscous fluid for very large Reynolds numbers,Doklady Akad. Nauk. sssr,30: 301, 1941.
Tatarski (V.I.), Wave Propagation in a turbulent medium.McGraw-Hill New-York, 1961.
Fried (D.L.), Optical resolution through a randomly inhomogeneous medium for very long and very short exposures.Journal of the Optical Society of America,56(10): 1372–1379, 1966.
Roddier (F.), Progress in optics XIX, chapter V:The effects of atmospheric turbulence in optical astronomy, pp. 281–376.E. Wolf,North-Holland publishing Company, Amsterdam, 1981.
Primot (J.), Application des techniques ďanalyse de surface ďonde à la restauration ďimages dégradées par la turbulence atmosphérique,PhD thesis, onera, Univ. Paris-Sud, Orsay, 1989.
Freischlad (K.), Koliopoulos (Ch), Wavefront reconstruction from noisy slope or difference data using the Discrete Fourier Transform.Proceedings of spie, 551: pp. 74–80, 1985.
Welsh (B.M.), Roggemann (M.C.), Signal-to-noise comparison of deconvolution from wave front sensing with traditional linear and speckle image reconstruction.Proceedings of spie, 2302: pp. 281–292, 1994.
Kern (P.), Optique adaptative et grands télescopes.PhD thesis, Univ. Paris VII, 1990.
Séchaud (M.), Hubin (N.), Brixon (L.), Jalin (R.), Foy (R.), Tallon (M.), Laser backscattered reference star for atmospheric wavefront disturbances measurement, In M.-H. Ulrich editor.Proceedings of ESO Conference on Very Large Telescopes and Their Instrumentation,30, pp. 705–714, E.S.O., Garching bei München, Allemagne, 1988.
Labeyrie (A.), High resolution techniques in optical astronomy.Progress in Optics, 14: pp. 49, 1976.
Korff (D.), Analysis of a method for obtaining near-diffraction-limited information in the presence of atmospheric turbulence.Journal of the Optical Society of America,63(8): pp. 971–980, 1973.
Knox (J.T.), Thompson (B.J.), Recovery of images from atmospherically degraded short exposure photographs.J. Astron.,193: L45-L48, 1974.
Bates (R.H.T.), Cady (F.M.), Towards true imaging by wideband speckle interferometry.Optics Communications,32(3): pp. 365–369, 1980.
Bondeau (C.), Bourennane (E.), Paindavoine (M.), A principal component analysis based method for the simulation of turbulence-degraded infrared image sequence.Annales des télécommunications,54, n° 5-6, pp. 324–330, 1999.
Beaumont (H.), Caractérisation de la turbulence atmosphérique et procédure ďamélioration des images pour des observations horizontales au-dessus de la mer,PhD thesis, Univ. Nice-Sophia Antipolis, 1996.
Ronfard (R.), Region-based strategies for active contour models,International journal of Computer Vision, 2: pp. 229–251, 1994.
Germain (O.), Réfrégier (Ph.), Optimal snake-based segmentation of a random luminance target on a spatially disjoint background,Optics Letters,21 (22): pp. 1845–1847, 1996.
Réfrégier (Ph.), Germain (O.), Gaidon (Th.), Optimal snake segmentation of target and background with independent Gamma density probabilities, application to speckled and preprocessed images.Optics Communications,137: pp. 382–388, 1997.
Réfrégier (Ph.),Germain (O.),Chesnaud (Ch.),Gaidon (Th.). Segmentation par contour actif probabiliste.Actes du 16e Colloque GRETSI sur le Traitement du Signal et des Images, pp. 1459–1462, Grenoble (France), 15-19 septembre 1997.
Chesnaud (Ch.),Réfrégier (Ph.). Optimal snake region based segmentation for different physical noise model and fast algorithm implementation.In psip’99 (Physics in Signal and Image Processing), pp. 3–11, Paris, 18-19 janvier 1999.
Chesnaud (Ch.), Pagé (V.), Réfrégier (Ph.). Robustness improvement of the statistically independent region snake-based segmentation method.Optics Letters,23: pp. 488–490, 1998.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Bourennane, E., Bondeau, C. & Paindavoine, M. Approches région et bayésienne pour la restauration ďimages dégradées par la turbulence atmosphérique. Ann. Télécommun. 56, 538–549 (2001). https://doi.org/10.1007/BF03008831
Received:
Accepted:
Issue Date:
DOI: https://doi.org/10.1007/BF03008831
Mots clés
- Restauration image
- Turbulence atmosphérique
- Processus stochastique
- Filtrage
- Méthode statistique
- Détection bord
- Propagation troposphérique
- Infrarouge
- Segmentation image