Research On Dehazing Method And Application Of Digital Image

Images of outdoor scenes, especially those taken in bad weather, are always degraded because of the existence of particles or water-droplets in the atmosphere.The degraded images always have low contrast and lose their faithful colors. Image dehazing can restore the contrast and faithful colors, so it can be applied in intelligent vehicles, remote sensing, video surveillance, computer vision, etc.This thesis first introduces the general dehazing physical model, analyzes the cause of degrading image quality, and then detailed introduces the dark channel prior(DCP) and dehazing algorithm based on DCP. So On this basis we introduce many other improved algorithms, which are based on median filtering, joint bilateral filtering, guided filtering.In 2010, Kaiming He first proposed the concept of guided filter, which has been quickly applied in many fields including image dehazing. Although guided filter is very fast due to Integral Image technique, the dehazing algorithm based on it can easily produce halo effect along edges with large depth jumps. For this problem, this thesis proposes an improved algorithm, which could remove the haze along edges with large depth jumps because of adoption of morphological opening-and-closing operation. Furthermore, it can filter the noise in far scene and boost the robustness of algorithm.Although dehazing algorithm based on DCP obtains a great effect on image dehazing, it has a high algorithm complexity and the process of refining transmission map is very time-consuming. Due to the high algorithm complexity of dehazing algorithm based on DCP, this thesis first proposed a dehazing algorithm based on Adaptive Local Noise Reduction Model. It can estimate the transmission map very fast by the adaptive local filter. It has function of “Brightness Adjustment” which can release the problem of underestimation of transmission map and make restored image more natural.