Image De-hazing techniques for Vision based applications - A survey

Abstract

Haze is defined as a poor condition described by an iridescent atmospheric appearance that reduces clarity and visibility. The main reason for this is lot of toxic elements like dust particles, smoke in the atmosphere scattering and absorbing sun light. This poor intelligibility causes various computer vision applications to fail, including intelligent transportation, video surveillance, element recognition, and in a method to perform operations on image to get better image. There is a problem in domain of image processing wherein image recovery by various degradations is a challenge. Pictures and videos taken in outdoor environments usually suffer from reduced contrast, faded colors and with reduced visibility due to airborne particles, which directly affect image quality. This can lead to problems recognizing objects captured in blurry or still images. Several images clean up techniques have been developed to solve this problem, each with their own strengths and weaknesses, but effective image recovery is daunting task. Recently, many learning-based methods (predictive analytics and natural language processing) have tried to overcome the shortcomings of mechanical representation of properties and alleviated the challenge of efficiently reconstructing images by spending with reduce cost and comparatively reduced time. This overview delves into latest techniques for imaging with no-fog. In addition, hardware execution of many real time dehaze methods have been methodically outlined by this paper. The study done in this paper paves a way for researches in image dehazing domain as-well-as will direct them for doing further enhancement on the basis of achievements done currently.