Single Image Dehazing Based on Dark Channel Prior with Optimal Scaling Factors

Abstract

Today, single image haze removal has attracted much attention in the community of image restoration. A popular type of approach is based on dark channel prior (DCP) because of its simplicity and effectiveness. One of the problems in the original DCP scheme is the fixed scaling factors for atmospheric light and transmittance. This paper attempts to replace them with optimal scaling factors obtained by the whale optimization algorithm (WOA). However, the WOA requires a reference to find the optimal scaling factors. To avoid using full reference in the optimization process, we adopt a half-reference that requires no ground-truth image. The proposed approach is justified by several examples and compared with the DCP scheme and a modified DCP scheme. The proposed approach shows its superiority over the comparison schemes in most of the examples, but has an overexposure or less dehazed problems in the rest of the examples. More research should be conducted on the proposed approach for better performance.