Fuzzy Logic Based Image Dehazing System

Image dehazing has emerged as a popular research field in recent years because of its' practical significance. It deals with restoration of visibility of images which are captured during hazy weather conditions. The degradation of visibility of these images occurs mainly due to scattering and attenuation of scene light caused by the presence of aerosol particles in the atmosphere. In this work, we have designed a novel image dehazing system to perform effective visibility restoration of images by exploiting their properties in multiple color spaces. The knowledge base of the designed system comprises of 125 novel Fuzzy Inference rules whose input fuzzy linguistic variables are chosen to be the selected parameters from multiple color spaces. The selection of parameters is carried out purely on experimental basis depending upon their relevance in detecting the intra-variation of visibility degradation among different regions of an image. Enhancement constant is considered as output fuzzy linguistic variable of the rules. Contrast enhancement of each pixel is performed using the unique enhancement constant value estimated based on its' characteristics. Comprehensive comparative analyses using several benchmark databases are carried out to validate the efficiency of the proposed system.