Research Of Fog-degraded Image Clearness Techniques

Under fog weather conditions, the contrast and color characters of the images captured by outdoor surveillance system are drastically degraded. In order to improve system's reliability and stability for object features detection, this paper focus the research on the fog-degraded image clearness techniques based on the analysis of image degradation mechanisms, and some novel/improved ideas and methods have been proposed mainly from two aspects which include image enhancement and image restoration.One of the main characteristics for fog-degraded images is that the local image contrast depends strongly on distance, that is to say, the level of contrast reduction increase exponentially with the distance from the camera to the object. As a result, conventional space invariant filtering techniques fail to adequately remove weather effects from images. Fog-degraded image clearness method based on moving mask proposed a novel idea for scene contrast enhancement, which does not need any scene depth information and avoid complicated atmospheric scattering model. However, computation complexity of the method is high and there is overlong time processing problem, aiming to solve this problem, an improved fog-degraded image enhancement algorithm is proposed in this paper and experimental results on fog-degraded images demonstrate that the proposed improved algorithm can gain better visual effect and solve the problem of long time processing to a certain extent.Considering that above improved method implements contrast enhancement by assuming all the pixels in moving mask with same depth, therefore, it can't deweather the fog-degraded images effectively when the scene in the mask has abrupt depth. In order to solve this problem, a novel fog-degraded image clearness algorithm based on depth region segmentation reprocessing is presented in this paper. The algorithm divides image scene into some different depth regions firstly, then enhancing the contrast in each depth region by histogram equalization function. By applying depth region segmentation reprocessing, this algorithm skillfully combines the important characteristic of fog-degraded images that the local image contrast depends strongly on distance and avoids exact depth information, therefore, efficiently decreases computing time and improve the visual effect.The images captured in fog weather usually have lower contrast and fuzzy object boundary, by combining the merit of fuzzy theory for uncertain information, this paper proposes a novel fog-degraded image clearness algorithm based on fuzzy contrast enhancement, which reasonably apply contrast exponential degrade law in fuzzy contrast measure modification, can perform fog-degraded image enhancement adaptively.For the problem of clearness restoration of fog-degraded images, it is of utmost importance to make a correct estimation of the parameters in the atmospheric scattering model, that is, the scattering coefficient and the depths of scene points. For the purpose of avoiding expensive hardware and the uncertainty from one fog-degraded image for distance, this paper proposed a new optics depth estimation method from two degraded images in different weather by reasonably combining the contrast exponential degrade law. Experimental results demonstrate satisfying scene restoration performance.In order to solve the clearness problem from single fog-degraded image, this paper improved the fog image degraded model and proposed a novel fog-degraded image restoration algorithm based on the improved model. By integrating the merit of genetic algorithm for searching global optimal parameters in this algorithm, the problem of fog-degraded images clearness restoration is transformed into the problem of optimization estimation for original undegraded image by maximizing the global contrast object function, in this way, the proposed algorithm can restore the object image as completely as possible in probability sense. Experimental results for single object image restoration gain satisfying visual effect.