Dehazing Algorithm Research For Single Hazy Image

Dehazing technology is an important research contents among image processing, the atomizing image is often accompanied by cross-color, contrast ratio reducing and image blurring. It will get into difficulties for studying and determining the object in scene, and serious influence usefulness of outdoor vision system. But the process of image degraded is a particularly complex, and existing dehazing models are not represent perfect as well as the know information is few of single image lead to existing dehazing algorithms are not satisfactory in changeful scenes. Consequently, studying the dehazing method of self-adaption and fast have very important theory and practice significance that based on the mechanization of image degraded in hazy weather. This paper have researched into the dehazing problems of single image as following:Scattering is a principal cause of image degradation. Based on the study for atmosphere scattering phenomenon, at first, analysising the wavelength's influence for scattering and studying the process and mechanization of image degradation, emphatic setting forth image degradation model's two submodels: incident light attenuation model and air light imaging model. Secondly, the paper make relatively exhaustive study for picture characteristics of hazy weather. This part establish the theoretical foundation for following study of image dehazing technology.Computing transmission is key for dehaizng algorithm of based on physical model. This paper present a self-adaption and fast dahazing algorithm base on scene radiance constraint, depending on existing dehazing algorithm need great space-time expenses when computing transmission. The new algorithm profits from the prior constraint condition—scene radiance constraint, the paper uses the prior to achieve the image dehziang by computing transmission with point-to-point and fast, and the amount of calculation reduce markedly as well as not produce the halo effect. The experimental results show image after dehazing not only can unveil the details and recover vivid colors even in heavily hazy regions, but also the visuality can obtain great raise. In additional, the new algorithm processing speed is fast and can satisfice the requirement of real-time for general engineering.This paper present an improved dehazing algorithm base on dark channel prior and scene radiance constraint, depending on existing dark channel prior dehazing algorithm need great space-time expenses and not process the sky-region. On the one hand, the paper uses guided image filtering in substitution for soft matting result in reducing space-time complexity and amount of calculation. On the other hand, the paper present an improved method for estimating the airlight, depending on existing insufficient for estimating airlight of the dark channel prior method. The improved method broaden the algorithm scope of application and strengthen the robustness. Furthermore, the biggest defect of dark channel prior will lose efficacy in sky-region. The paper combines the superiority of scene radiance constraint and dark channel prior overcome the defect that lose efficacy in sky-region. The experimental results show image after dehazing not only can unveil the details and recover vivid colors even in heavily hazy regions, but also the visuality can obtain great raise. In additional, the new algorithm can process the sky-region.