Multiple scattering model based single image dehazing

Single image dehazing has been a challenging problem due to its ill-posed nature. Most of the existing dehazing algorithms are based on single atmospheric scattering model. However, this model is not capable of explaining complex effects of weather on images as multiple scattering occurs in real world. Existing multiple scattering models are widely utilized for simulating various weather condition; however, only a few algorithms are applied for dehazing. In this paper, we propose an image degradation model based on multiple scattering for image haze removal. We improve the widely utilized dichromatic model with atmospheric point spread function. Applying deconvolution to the dehazed image with the estimated blur kernel, the inherent scene radiance is restored. Experimental results with a benchmark dataset demonstrate that our approach significantly improves the contrast and sharpness of the hazy image.