Research On Image Dehazing Algorithm Based On Dark Channel And Color Attenuation Prior

Image dehazing is one of the research hotspots in the field of image processing and computer vision.This technology can enhance the constrast of haze image,improve the visibility of the image scene and the visual effect of the image,and reduce or even eliminate the effects of aerosols in the atmosphere on image by using specific methods.It has a high application value in intelligent vehicle,outdoor monitoring system,remote sensing,military reconnaissance and other fields.Image dehazing based on atmospheric scattering physical model is an important image dehazing method,in which it uses inverse solution of computing the atmospheric scattering model under certain prior information to obtain the haze removal image.Based on the analysis of the atmospheric scattering model,the image dehazing algorithm based on dark channel prior and the image dehazing algorithm based on color attenuation prior are studied in detail in this thesis.Because the atmospheric light value obtained by the dark channel algorithm is easily affected by the white object in the foreground of the image,the average value of all points that meet the conditions is selected as the global atmospheric light intensity value to avoid the restoed image too dark or distorted and improve the reliability of atmospheric light intensity value.As the dark channel prior method is not suitable for the sky or the large white area in the image,the image transmission rate is modified to reduce the distortion phenomenon of the areas.According to the advantages and disadvantages of the image dehazing algorithm based on dark channel prior and the image dehazing algorithm based on color attenuation prior,an image dehazing algorithm based on dark channel prior and color attenuation prior is proposed.Firstly,the image depth information is obtained based on attenuation prior.Next,the pixels of the top 0.1% of the brightness values are chosen in the depth image and the corresponding pixels in the original haze image are used to calculate their average value that is used as atmospheric light value.Then,the depth information is used to segment the sky and distant dense haze of the haze image.As for the image foreground,the dark channel prior method is used to calculate the transmission and the minimum information loss is used to constrain it to obtain optimal transmission.As for the sky and distant dense haze,the transmission is estimated using the depth information.The guided filter is used to refine transmission to obtain the optimal transmission.Finally,the atmospheric light value and the optimal transmission are obtained to solve the atmospheric scattering model and restore the image.The experimental results show that the proposed method can effectively restore haze images with sky and other white objects or distant dense haze,the dehazed images with high contrast and rich detail informations...