Vision Enhancement of Single Foggy Image using CNN

Abstract

Fog removal from a single foggy scene is a tedious piece of work. Some of the existing methodologies are based on various constraints and assumptions to evaluate fog free image or defog image. Recent researchers have applied deep neural network algorithms to calculate defog image. Motivated by the recent works, this paper aims to estimate fog free image entrenched on Convolutional Neural Network (CNN). Different haze relevant features are learned using deep architecture of CNN. Using these extracted features, final defog image has been estimated. Results are compared with the existing methodologies for analysis purpose as well.