Research On Image Haze Removal Under The Foggy Weather Condition

In recent years, with the serious environmental pollution, the visibility is seriously reduced by the haze, which induces people make wrong judgments. The cause of the error is that people have no access to accurate information. More worrying is the lack of information could incur disasters. The atmosphere will have abundant suspended particle (e.g. the drop of water, aerosol et al.) in foggy days, which attenuates the luminance of the ray due to the scattering phenomenon in the process of transmission through the foggy atmosphere. Therefore, the image contrast will be degraded, the details will be blurred, the most features of the image will be covered, the color fidelity also will be reduced. These degradations of the image features could directly impact the performance of the outdoor monitoring system, the traffic and transportation system, remote sensing system et al.. Therefore, this paper makes a deep research on the image haze removal.Firstly, the paper researches the atmosphere scattering model and the dark channel prior model. The relative merits of the dark channel prior model are analyzed. Due to the restored image which is handled by the dark channel prior model in the condition of heavy fog appearing the phenomenon of the halo, this paper proposes an optimization algorithm based Guided Filter whose edges are optimized. The phenomenon of the halo can be removed by segmenting the border of the mini-map to re-estimating the transitivity in the area whose depth of field is severely changed. The experimental results showed that the proposed algorithm is robust.Secondly, because Dark Channel Prior Algorithm cannot improve brightness and details of low illumination images, and the image enhancement technique can well improve contrast and brightness, this paper researches the histogram equalization, homomorphic filtering, Retinex et al.. After enhanced by image enhancement techniques, haze image which has low illumination have higher contrast and brightness, while color is distortion and gray. Hence, this paper proposes a haze image with low illumination imaging model, to show how the fog and low light effects the scene imaging. At this basement, an attenuation component enhancement algorithm is proposed. This algorithm can enhances the details of dark areas as well as removes the fog.In the end, in view of the traditional Retinex algorithm cannot effectively restore details under the condition of image with uniform fog, an adaptive local Retinex enhancement algorithm is proposed. In order to improve processing efficiency, propose a Retinx enhancement based on HSV color space for foggy and fogless image. At last, combined with local Retinex, a local Retinex enhancement based on HSV color space is proposed, not only enhance sharpness and restore color, but also improve execution efficiency.