Research On Image And Video Dehazing

With the increasing improvement of image processing and optical imaging technology,outdoor visual systems obtain images with higher resolution and are widely applied to daily life,which bring great convenience for outdoor photography,traffic detection,video surveillance and so on.However,due to the scattering process of suspended particles,the images and videos obtained in hazy days are usually severely attenuated with lower contrast,color distortion and loss of details.In order to reduce the effect of haze on outdoor visual systems and collect clear and useful images in a stable way,studying efficient image dehazing technology to improve the contrast of fog-degraded image and restore the clarity of the sight has improtant application value in meeting people's visual needs.Firstly,this paper studies the degradation mechanism of images obtained in foggy days and summarizes the classical dehazing algorithms.The theory of dark channel prior has better dehazing performance.However,when processing the sky region of outdoor images,the algorithm obtains lower transmissivity value,resulting in halo artifact and color distortion.This paper proposes an improved image dehazing technology based on image region segmentation.The proposed method utilizes improved recursive Otsu segmentation to accurately separate the sky region.Then it utilizes the normalized gray values of the dark channel map to optimize the transmissivity of the segmented region.Finally,the haze-free image can be restored by the atmospheric scattering model.Compared with other algorithms,the proposed method can restore the details of images in a more accurate and natural way and solve the problem of halo artifact and color distortion,especially that of boundary and sky region.Secondly,the atmosphere scattering model fails to meet the need of real-time video processing,basing on the image enhancement technology of non-physical model,this paper analyses the characteristics of histogram redistribution of foggy images by using histogram equalization and proposes a kind of improved video dehazing method of Contrast Limited Adaptive Histogram Equalization.In view of space transformation,the method transforms the color information of foggy images from RGB color space to HSI color.space,holding component H constant and conducting piecewise exponential transformation on component S to increase the saturation.It conducts CLAHE and bilinear interpolation on component I to limit noise amplification and over-enhancing of local contrast,improve the operating speed of the algorithm and ensure the real-time performance and effectiveness of video processing.Verified by mulit-set of experiments through stimulation,the method has distinct advantages in many aspects such as information entropy,foggy image average gradient and PSNR.It has natural dehazing visual effects and satisfies the need of real-time video processing.