RESTAURACIÓN DE LA VISIÓN DIGITAL EN AMBIENTES CORROMPIDOS POR DEGRADACIONES NEBULOSAS

Abstract

Restoration of images which have been corrupted by foggy degradations such as the presence of rain, fog or mist, has received an increasing research interest in recent years. This is because recovery of a clear visual perception from a foggy scene has great impact on the decision making in environments where a bad decision can bring fatal consequences in critical applications such as vehicle navigation. The main effect of a foggy degradation in an observed image is the corruption of the intensity levels that the opto-electronic sensors capture and which are reflected by the objects present in the scene. As the result, the visual contrast in the observed scene is severely affected causing that distinguishing of present objects in the scene become a very difficult task. Image restoration in foggy environments is a difficult problem because foggy degradations can be modeled by a highly correlated process which statistical parameters can vary over time and moreover, the computational load that is required for the restoration process can be very intensive. In this thesis we propose two computational algorithms for restoration of images corrupted by foggy degradations, which make use of the physical image formation models. The proposed algorithms are based on the use of a sliding-window, in which by means of the physical model of the foggy degradation the local scene-depth is estimated in each position, to get the restored signal. Several computer simulation results obtained with the proposed algorithms are presented, discussed and compared, with different methods for image restoration in foggy environments.