Haze editing with natural transmission

Significant efforts have been devoted to haze removal of outdoor scenic images and haze simulation of virtual scenes. However, fewworks focus on editing (increasing and decreasing) haze effects, which are common outdoor photography on real world images. In this paper, we present a dark channel prior-based transmission model that can explicitly formulates aerial perspective implying human perception on natural haze. We introduce maximum visibility as a parameter into the transmission model, so that we are able to naturally edit the amount of haze in an image by tuning this parameter with a physical interpretation. Additionally, we derive color correction and sky compensation from the transmission model, which improves the image quality for haze editing. Experimental results demonstrate the ability of the proposed method to generate images with various amounts of haze in a natural and efficient manner. Comparisons with the traditional algorithms on haze removal show the performance of the proposed algorithm in terms of two objective metrics that evaluate the visibility and fidelity of the restored images.