Driver Assistance System based on Real-time Dehazing

Abstract

Weather conditions such as Haze, Fog reduce visi-bility making it difficult to observe the surrounding environment. This makes drivers prone to accidents in such situations. Most of the existing methods for dehazing use models that work upon single-image dehazing. Dehazing in real-time proves a much bigger challenge as it demands efficient handling of the constant stream of frames, as well as a lightweight dehazing algorithm. Further, there is the need for some decision-making parameter that determines whether dehazing is needed. This paper provides a novel framework that works at dehazing in real-time using Dark Channel Prior(DCP), and is able to work on minimal hardware such as smartphones.