Research On Image Dehazing Algorithms

In recent years,the heavy haze emerging in our country frequently has not only posed a risk to human health,but also seriously affected the imaging quality of outdoor vision system.Meanwhile,it limits the contrast of the images and causes color distortion,so that the image cannot achieve the satisfactory visual effects.Due to the effects of haze,it is hard to make full use of the systems of target identification and tracking,visual surveillance,military reconnaissance and remote sensing imaging.Therefore,image haze removal is an urgent problem to be solved in the field of computer vision.This paper proposes a new single image dehazing algorithm based on dark channel fusion and wiener filtering,which overcomes the defaults of higher time complexity and inaccurate atmospheric estimation of the dehazing method based on dark channel prior.Combining the advantage of patch-based dark channel prior and pixel-based dark channel prior,a coarse transmission map is estimated by fusion of those two dark channels using the wavelet transform,and the adaptive wiener filter is introduced to further refine the transmission map.Simultaneously,this paper proposes the weighted quadtree subdivision to estimate atmospheric light.Experimental results show that our algorithm can get good visual effects and lower the time complexity to meet the real-time application.This paper also analyzes and researches the characteristics of haze image in inhomogeneous atmosphere.A new image dehazing algorithm based on sky region segmentation is proposed,which overcomes the defects of inaccurate estimation of transmissions in sky area.A transmission map is estimated based on boundary constraint and weighted TV.With the low gradient and high intensity of the sky,the transmission can be refined based on the sky region segmentation,this method achieves the good recovery effect while retaining the details in sky area.