The Research On Image Clearness Method In Haze Weather

Because of the development of the computer technology,communication technology,broadcast television technology and the reduction of image-process-facility's price, the application of the outdoor visual system( such as outdoor safety monitor and road monitor )is used more and more widely. The demands of system's adaptability are raised, for example: the outside visual system can still work, even in bad weather conditions. This paper studies on the methods of foggy image clearness in the haze weather condition. The purpose of this paper is to work out a new algorithm. This new algorithm can not only enhance the fog-degraded images while having a good time complexity but can also give help to image clearness on the other inclement weather conditions. The main contents of the thesis are as follows:At present, there are two kind of technique for fog-degraded image's enhancement.First one is the model-based images recover technique. From the notion of physics reason this method is to model function of the atmosphere scatter and analyses it.Second one is at the side of the image process method to enhance image contrast.This paper studies on second one, meaning using enhancement algorithm to enhance the image's contrast. The technology of foggy image enhancement can classify to two classes. The first one is based on histogram equalization algorithm. Another one is the Retinex algorithm. This paper is study on this two algorithm, and improve the POSHE and center/surround Retinex algorithm.1.As the foggy image's gray levels are only a little and collected to a small range of gray levels, while adopting traditional histogram equalization algorithm to enhance the foggy image, the result's macrocosm effect is good, but local detail is failed to enhanced. And some local HE such as non-overlap HE,POSHE still exist defects that are noise amplification and block effect. Because of these defects this paper first improves the traditional HE by a parabola function and then combine with sub-block overlap operate to improve the POSHE algorithm. The improved algorithm adjusts the brightness of the enhanced images, and it basically eliminates the block effect, while the local details get effectively enhanced, without noise amplification.2. This paper analyzes different Retinex algorithms and improves the SSR and MSR algorithms, because of their defects in enhancing the foggy images. First let the reflect image that is got by the SSR algorithm to transform via the Sigmoid function to extend its gray range. Then use wavelet decomposition's multi-resolve to integrate the different SSR's results to replace the liner weighted to improve the MSR algorithm. The improved MSR algorithm compare with traditional MSR algorithm can enhance details more effectively.