General Participative Media Single Image Restoration

This paper describes a method to restore degraded images captured in general participative media — fog, turbid water, sand storm, etc. To obtain generality, we, first, propose a novel interpretation of the participative media image formation by considering the color variation of the media. Second, we introduce that joining different image priors is an effective alternative for image restoration. The proposed method contains a Composite Prior supported by statistics collected on both haze-free and degraded participative environment images. The key of the method is joining two complementary measures — local contrast and color. The results presented for a variety of underwater and haze images demonstrate the power of the method. Moreover, we showed the potential of our method using a special dataset for which a reference haze-free image is available for comparison.