Single image dehazing algorithm based on scattering blur prior

Fog or haze inevitably leads to severe degradation of optical images, manifesting as color distortion and reduced contrast. Fortunately, numerous studies have offered profound insights and effective solutions to this problem. However, most of the existing methods are ineffective when the scattering media are dense. Here, we propose a method named Scattering Blur Prior (SBP) to address the challenge of restoring a severely degraded image captured in dense haze. A crucial innovation in mechanism is a prior assumption that the blurring effect in dense fog stems from the back-scattering-induced contrast compression and forward-scattering-induced spatial blurring. Additionally, the latter effect is always deliberately ignored in other methods, because it is difficult to obtain the multi − scattering target light from a single image. Further, a key step in the technique is a convolution − based blurring calculation aimed at successfully separating the blurring effects of multi-scattering target light from the blurry background of atmospheric scattering light. In contrast to state-of-the-art methods, the proposed approach achieves superior restoration performance—especially under extreme dense fog—across both qualitative visual assessment and quantitative metrics. Most importantly, the SBP method has been widely proven to successfully reconstruct the images captured in extreme scenarios with dense fog, haze, or smoke. These achievements highlight the great potential of our method as a robust pre − processing module for real − world low − level vision pipelines.