Study On Image Defogging Based On Dark Color Prior Theory

Image defogging algorithm aims to enhance image’s contrast, restore image’s color feature and some important details information, used to make outdoor surveillance system work normally and stably, reduce the occurrence of traffic accidents and traffic jam in fog weather. Currently, because of the satisfactory defogging effects, image defogging method based on dark channel prior theory has been widely concerned, but this algorithm has two disadvantages: firstly, the time and space complexity is too high, which makes it can’t meet the real-time requirements; secondly, dark channel theory will out of work in light color area, which will lead to a overly enhanced restoration in these area.Considering to increase the practicality of the defogging algorithm based on dark channel prior theory, the paper took how to reduce the algorithm’s complexity and extend the application scope of the dark channel prior theory as the research key, made a deeply research on defogging method based on dark channel prior. Main achievements of the paper include:(1) The paper proposed a novel improved image clearness method based on dark channel prior. First of all, target and background at each image block are segmented self-adaptively, and more accurate dark channel in each image block is obtained at the region of target; then wavelet threshold edge-preserving algorithm and wavelet enhancement algorithm are used to repair the inaccurate brighter dark channel in smooth area; At last, the fog-free image is obtained by using the physical model. The experiment results show that this method can restore the fogging image effectively and reduce the time complexity.( 2) Combined with edge detection algorithm, another improved image clearness method based on dark channel prior is proposed to solve above insufficient. Firstly, dark color image was obtained and decomposed by wavelet; secondly, smooth portion come from wavelet decomposition was used as the target image, in which global atmospheric optical A and the dark channel value were calculated, then, edge information of target image was extracted and dilated; thirdly, according to the expanded binary edge image, the dark channel value in edges and that in smooth parts are calculated and optimized, respectively; at last,according to the transmission map calculated using the optimized dark channel values,the fog-free image is obtained by using the physical model. The experiment results show that this method can restore the fogging images effectively on both of visual effect and complexity effect.(3) At the same time, consider adjusting the overly enhanced restoration, which led by the dark channel theory out of work, in light color area, an improved algorithm based self-adaptive threshold mechanism is proposed used to enhance the transmission map in these light color areas.At the end of paper, there is a sum-up for all research work and the future research direction about image de-fogging is indicated.