Research On Nature Image Dehazing

Images captured in fog, mist and haze weather conditions often suffer from bad visibility and lose contrast and color fidelity, which result in outdoor imaging system fail to extract features effectively in the scene and greatly affect the effectiveness of the relevant system. Hence, research on nature image dehazing algorithm in the field of computer vision and image processing gradually draw the attention of many scholars. This paper focus mainly on hazy images dehazing algorithm based on atmospheric scattering model. The main contributions in this paper are shown as follow:An improved nature image dehazing algorithm based on guided filter is proposed, which employs the atmospheric veil to refine the original transmission map and can efficiently overcome the issue that the area are not smooth with the same depth in the traditional algorithm. First, in the proposed algorithm, the dark channel prior based on visual perception is utilized to estimate the raw transmission map. Second, the hazy image is applied to estimate the guided image that consists with raw transmission as the input of the guided filter. Meanwhile, the refined transmission that reflects the scene depth information and others parameters are obtained. Finally, the restoration of the original image can be performed by solving the haze degraded model. Experimental results illustrate that our method avoids the loss of detail information as well as remaining fog existing in traditional algorithms and achieves richer details, clearer structures, more faithful color.A natural image dehazing algorithm based on the image gradient minimizing using L0-norm is proposed. First, the proposed method obtains transmission based on a prior of the boundary constraints in the scene. Second, the transmission of the scene is evaluated by builting the boundary constraint problem into an optimization problem combined with the gradient minimization L0 smoothing, and then the hazeoff image is got using foggy image degradation model. The proposed algorithm is able to effectively keep information at the edges of objects and obtains more distinct effect in this region when optimizing the rough transmission. Simultaneously, it also improves the dehazed results on the region where dark channel prior fails. Experimental results show that the dehazed image has richer details, clearer structures and better visual effects. Furthermore, an objective quality assessment method is applied to evaluate the mainly distortions existing in the results of several dehazing algorithms. Experimental results show that both the our proposed dehazing algorithms all have a desirable effect on details recovery, colour revivification and structural sharpness.