Image Dehazing Algorithm Based on Particle Swarm Optimization for Sky Region Segmentation

Abstract

The well-known Dark Channel Prior (DCP) dehazing algorithm does not work well in the sky area, thus, we propose an image dehazing method for segmented sky regions based on Particle Swarm Optimization (PSO). First, we apply PSO to the Otsu segmentation method to accurately segment sky and non-sky regions in the hazy image. Second, the transmission of non-sky regions is derived by DCP. A transmission compensation method is proposed to obtain the transmission for the sky regions. The transmission is then refined by gradient-domain guided filtering. Finally, the restored haze-free image is obtained through the Atmospheric Scattering Model (ASM). Quantitative and qualitative experiments show that the proposed method has a better dehazing effect, especially for hazy images with the sky areas.