Research On Image Dehazing Enhancement Methods Based On Dark Channel Prior

Image dehazing technology is a cross-cutting issue in the field of image processing and computer application.When the weather conditions are not good,the suspended particles in the air have a negative impact on the propagation of light,which results in the low image quality obtained by the capture device and the great inconvenience to tasks such as target retrieval and behavior tracking in the field of computer vision.Based on the atmospheric scattering model,this thesis analyzes the causes of image degradation under haze weather and proposes two algorithms based on dark channel dehazing,by which good dehazing results can be obtained.The main works of this thesis is stated as follows:(1)An image dehazing algorithm based on optimized transmittance and semi-inverse is proposed.Aiming at the failure of the dark channel at the edge of the image,the transmittance correction mechanism is used to eliminate the band phenomenon at the edge.And then,the atmospheric scattering model and the improved semi-inverse method are used to obtain the overall illumination value of the atmosphere.Finally,according to the phenomenon that the color of the whole picture becomes dim after dehazing,the image contrast algorithm based on wavelet transform is used to process the image after dehazing to improve the visual effect.(2)A fast image defogging method based on sky segmentation is proposed.In order to make the transmittance value more accurate and better recovery of the sky region,instead of the minimum filter,the median filter is firstly used to obtain the initial transmittance,and the Canny operator is also introduced to segment the sky region and the non-sky region to correct the transmittance value in the non-sky area.Then,based on the atmospheric model,a rapid estimation of the atmospheric light value method is adopted,which can reduce the computational burden and make the atmospheric light value estimation more accurate.Finally,in order to improve the dimness of the picture after defogging,the color of the image is enhanced by the contrast limited adaptive histogram equalization method.The results on the atomized image set show that the proposed two image dehazing methods in this thesis are obviously superior to the original ones.In addition,compared with some advanced algorithms,the results by the fast image dehazing algorithm based on sky segmentation is better.