Study On Image Dehazing Based On Image Fusion

In most areas of northern China,large area of rain,snow and haze are often founded during autumn and winter season,and the bad weather conditions seriously affect the daily production and life of the residents.At the same time,the image contrast ratio reduce,the image blur,the details of the image are lost,and the thick fog also restricts the outdoor monitoring and setting.The use of the preparation.Therefore,the restoration of high-quality clear images is of great practical significance.In this paper,the main content of the algorithm is following:Based on the theory of dark channel prior and color attenuation prior,two kinds of image fusion fogging methods are proposed respectively.Experimental results show that the two methods greatly improve the image dehazing result.The first image fusion fog dehazing algorithm proposed in this paper is based on dark channel prior algorithm to fuse fog removal method.Dark channel prior algorithm is the most close to the fog essence of the fog removal method,but this method can cause the color of the dehazing image dark,and in the sky and other bright regions can appear color distortion.In this paper,a tolerance parameter is introduced to overcome the distortion of dark channel prior in the sky area to abtain the first image be used for fuse.Then through analyse the H,S,I three color components before and after dehazing.S and I component CLAHE enhancement respectively,then converted back to the RGB color space to get another dehazing image of high brightness,and high saturation,finally let the two images fuse to get dehaing image.The results show that the method is better than the dark channel prior algorithm.The second image fusion dehazing methods are improved base on color attenuation prior.First,a parameter model of the image transmissivity map,the image saturation and the brightness of the image is established by the machine learning gradient descent method.Then the inversion graph of the gray image of the foggy image is filtered.Finally,the two transmissivity maps are fused by the method of LA's Pyramid fusion to get the transmissivity map after the fusion.then use the transmittance information Substitution the physical model of atmospheric scattering.Experimental results show that the algorithm has less time overhead compared with other classical image dehazing algorithms,and the natural color and visual effect of the image after fog removal are good.