Transparency-based Vision through Haze

Lightness illusions are such special visual phenomena that give solutions to lightness constancy and help for further image processing tasks such as haze removal. Based on the research of haze illusion, transparency theory and basic atmosphere transfer functions, we propose the tATF model to dehaze. With the idea of filtering, three parameters in the model are calculated respectively; especially the additional transparency term t is obtained by perceptual transparency experiments. Furthermore, a new quantitative standard haziness is naturally presented for evaluating the dehazing effect. Experimental results illustrate that the proposed method effectively yields haze-free images with better objective evaluation values.