Single Image Dehazing In Inhomogeneous Atmosphere With The Influence Of Intense Light Source

In recent years, as air pollution and air quality continue to deteriorate, hazy wea-ther occurs frequently across the country. Due to absorption and scattering effects of suspended particles in hazy atmosphere, the images captured by the device usually have bad quality in contrast and fidelity. Therefore, there is practical significance in research on image dehazing to ensure that outdoor imaging system can work more effectively in bad weather condition.Since the reason of degradation of hazy image is complex and part of the scene information is always lost, it is difficult to restore the scene from a single hazy image. At present, the degradation of model, that most current available dehazing algorithm rely on, is a simplified model in condition of homogeneous atmosphere. In practice, however, many cases will result in inhomogeneous atmosphere condition, such as, the slant observation, influence of intense light source,etc,which will cause that the illumi-nation and optical properties in the sight path is non-uniform. Further more, wether the degradation model can accurately reflect the hazy imaging principle is crucial to the dehazing results.In this paper, we will focus on the inhomogeneous condition that influenced by the intense light source. First, starting from the transport theory of atmosphere, and based on the image degradation model in homogeneous, we establish a degradation model for the inhomogeneous atmosphere via analyzing the illumination and optical properties influenced by the intense light source in sight path. Then, based on the proposed model, the dehazing algorithm, using Dark Channel Prior, for the inhomo- geneous atmosphere influenced by the intense light source is introduced in detail. Finally, we use our dehazing algorithm to processing the nature outdoor images download form the website, and take comparison of experimental results with other typical methods. Experimental results show that the proposed algorithm is effective for the hazy image in inhomogeneous atmosphere influenced by intense light source.