A modified prior-based single-image dehazing method

Images acquired outdoor suffer from degradation in bad weather such as haze and fog, which heavily drops the visual quality. Recently, many dehazing algorithms have been proposed. However, prior-based methods tend to introduce artifacts and magnify noise. This paper provides a modified prior-based dehazing pipeline to solve some drawbacks of local prior-based dehazing methods. We translate the accurate transmission map estimation into the parameter estimation problem of a modified global fusion model. Based on our model, oversaturated outputs will be remedied. Experiments demonstrate that our method can handle many conditions and obtain wonderful results from real-world haze images stably compared with other methods.