Development of Image Dehazing Algorithm

In recent years, image dehazing algorithm has been widely used in the preprocessing of advanced computer tasks, such as target tracking and pattern recognition. The traditional image dehazing algorithm, which relies on the atmospheric scattering model, realizes fast dehazing by means of image enhancement and image restoration, but its application scope is very limited. In the data-driven mode, a large number of researches directly establish the mapping relationship between haze image and clear image, and obtain better effect on the synthetic dataset. However, how to make the dehazing network more suitable for the real haze dehazing is still an urgent problem to be solved. Therefore, weak supervision is applied in this field.