Distant horizontal ground observation: atmospheric perturbation simulation and image restoration

Atmospheric perturbation is certainly the most important degradation when observing a scene horizontally at a distance of several kilometers. Because such images with short time exposure are expensive to obtain, they have to be simulated. The goal of this paper is to adapt simulation and restoration of images perturbated by atmospheric turbulence to the case of horizontal ground observation at a large distance and with a short exposure time. The first part of this paper is dedicated to the simulation process. First physical characteristics of atmospheric turbulence are exposed. Then the simulation algorithm is detailed and is adapted to both the anisoplanatic and isoplanatic cases. Simulations have been made with different ratios D/r0 corresponding to different turbulence intensity values at two wavelengths: one in the visible domain and another in the near infrared domain. Moreover the generation of image sequences where the temporal evolution of the atmospheric turbulence for a fixed scene (i.e. without camera motion) is taken into account, is studied. Indeed the temporal atmospheric turbulence evolution is a key element to be considered when recovering the original scene is desired. The method to obtain such a sequence is explained and an extract of simulated sequence is given and commented. In the second part of this paper, four different classical restoration methods have been tested on these image sets, which can only be used with single frames and need also some knowledge about the original image. To cope with this problem a more recent method has been tried: it is especially adapted to image sequences degraded by atmospheric turbulence. Comparisons have been made between the different obtained results. To conclude some perspectives are given.