Fast airlight estimation algorithm in dark channel prior for image dehazing applications

Abstract

The image dehazing can recover the hazed image back to the nearly haze-free one. Among sophisticated image dehazing algorithms, the dark channel prior (DCP) method is more efficient owing to its effectiveness. However, the airlight estimation of DCP causes a considerable computation burden. To overcome this problem, a fast airlight estimation algorithm for DCP is proposed in this wok. The experiment result demonstrates that the estimation error rate is as minor as 1.07%, and the computation load can be reduced by 92.82%. This work can save a great amount of computation burden, while keeping a highly accurate airlight estimation for DCP.