Appraise of Deep Learning and Image Processing based Single Image Dehazing Algorithms

摘要

The atmospheric particles tend to either absorb or scatter the light rays. This phenomenon is known as the haze by which light and other natural characteristics are obstructed. Image dehazing is an image enhancement approach that refers to procedures that attempt to eliminate the haze amount from a hazy image. The single image dehazing is a dehazing approach that uses a single reference image for dehazing. Dehazing methods are majorly classified into two main categories, one of them being prior based where most of the methods are linear and calculation heavy which removes haze on the most prominent features. While the other one is data-oriented which understands the image features and extracts them and converts them either in some useful maps or in some feature maps. Both methods are efficient in their view and both have some limitations in them. This paper presents a review of several single image haze removal methods. The review covers different prior-based and data-oriented approaches for image haze removal.