The Research Of Clearing Images Dehazing Algorithm Based On Dark Channel Prior

In bad weather conditions, such as the fog, haze and other environmental factors,imaging system suffers from scattering and refracting causes by the tiny particles in atmosphere. So this leads to decrease of the color saturation and contrast ratio of the obtained images. As a result, important detail information of the images misses. This makes against feature extraction and identification of images. And it severely restricts the application of outdoor surveillance system and detection system. In order to obtain clearer images of enhancing clarity and improving identification, the thesis delves into the theory of the advanced dark channel prior dehazing algorithm on the basis of analysis the haze images of physical model. Dehazing utilizing the theory of dark channel prior is simple and effective,restoring the advantages of good effect. But there are also disadvantages of weak dehazing ability and color distortion phenomenon in bright area. Aiming at the shortage of dark channel prior algorithm, the thesis proposes two kinds of new improvement algorithms. The main research work and content are as follows:Degraded images can be divided into images containing bright area and images without bright area. The thesis proposes algorithm of compensation and dehazing algorithm of adaptive threshold based on guidance filtering.(1)The thesis analyses the reasons of block and white edge effect. The dark channel images are compensated through median filtering of Gibson in degraded images without bright area. The guided filter that used the minimum image as guidance image optimizes the new dark channel images. It can estimate transmittance indirectly and improve blackspot phenomenon caused by overvalued local dark channel due to the median filter.(2)Through in-depth analysis of the characteristics of atmospheric veil, the atmospheric veil can be gotten in the way of guided filtering the minimum image in containing bright area of degraded images. And the guidance image is the local maximum value of minimum image. Meanwhile, contrast ratio can be obtained by method of absolute error between the atmospheric veil and the minimum image. The aim is to discriminate fog area and the overvalued close shot area itself and get more accurate atmospheric veil. Combining tolerance mechanism with adaptive threshold which is the layering statistics of pixels of bright area can modify the too small transmittance. The purpose is to improve color distortion phenomenon in bright area.Meanwhile, two types of images are processed in different methods in evaluating the atmosphere light. The atmosphere light of degraded images without bright area still uses dark channel prior theory. In order to obtain more accurate estimate of the atmosphere light in degraded images containing bright area, it needs to add adaptive threshold to ascertain the accurate location of the sky.Experimental results show that the obtained dehazing images utilizing algorithm of compensation are clearer and have better color saturation. Meanwhile, the detail area has the advantages of higher contrast ratio, and eliminates the blackspot effect. And the bright area in obtained images utilizing algorithm of adaptive threshold based on guidance filtering is clear and natural at the same time close shot area in dehazing images has good effect. The overall contrast ratio and sharpness of images and computational speed are greatly improved.