Contrast based background and foreground channel prior for single image dehazing

Abstract

ABSTRACT To reduce haze in an image, a contrast-based background and foreground channel prior (BFCP) is proposed in this work. The haze will affect both colour and contrast, and the majority of haze reduction techniques rely on the dark channel prior (DCP) as well as the dark and bright channel priors (DBCP). Even though the DCP and DBCP produced accurate results, they provide a less precise estimate of the haze depth because they do not take into account the local contrast and spikes in the haze image histogram. The goal of this BFCP is to improve the estimation of the transmission map and to correct the spikes in the histogram of the hazed image. By utilizing contrast differences in the background and foreground, BFCP improves transmission map estimates. This method is tested on images of (i) indoor haze, (ii) outdoor haze, (iii) dense haze, (iv) night haze, and (v) fog.