An Image Dehazing Algorithm Based on Improved Atmospheric Scattering Model

The scene incident light is always assumed to be a global constant when the traditional atmos- pheric scattering model is used in the image dehazing. However, the assumption is unreasonable. Thus, an image dehazing method based on an improved atmospheric scattering model is put forward. Firstly, a hazy image was classified into different scenes by the brightness channel prior and the fuzzy clustering. Secondly, the incident light of each scene was estimated. Then, the scene structure was estimated by the optical radia- tion characteristic. After that, the transmission can be calculated by the haze concentration estimation model. Finally, the hazy image is recovered with the improved atmospheric scattering model. A large number of comparative experimental results show that the proposed algorithm can recover a more detailed, clear and natural haze-free images. Besides, the calculation speed is relatively fast which can meet the real-time re- quirements of general engineering.