Single Image Haze Removal Based on a Simple Additive Model With Haze Smoothness Prior

Single image haze removal, which is to recover the clear version of a hazy image, challenging task in computer vision. In this paper, an additive model proposed approximate formation process. contrast with traditional optical model, it regards as layer clean image. The thus avoids estimating medium transmission rate and global atmospherical light. addition, based on critical observation that changes gradually smoothly across smoothness prior constrain model. This assumes much smoother than Benefiting from prior, we can directly separate single Experimental results comparisons synthetic images real-world demonstrate method outperforms state-of-the-art removal algorithms.