Research On Dehazing Algorithm Based On Dark Channel Prior Theory

The quality of images collected by imaging equipment in haze weather will degrade to different degrees,which seriously affects the normal operation of computer vision system.Therefore,the dehazing algorithm for foggy images is an important research direction.Focusing on fog image dehazing algorithm,the fog image degradation model,image dehazing algorithm based on dark channel prior theory,improved image dehazing algorithm based on dark channel theory and fast image dehazing algorithm based on Gaussian blur are studied in this thesis.The specific content is described as follows:1.The formation principle of haze weather and its influence on image degradation are analyzed.The atmospheric scattering model is introduced as a foggy image degradation model.The model consists of two parts: incident light attenuation model and atmospheric light enhancement model.The image quality evaluation method is introduced to evaluate the image quality.2.The image dehazing algorithm based on dark channel prior theory is discussed,including dark channel prior theory,transmittance estimation,soft matting to optimize transmittance,estimation of atmospheric light value and image restoration.The advantages and disadvantages of the algorithm are analyzed by simulation experiment.3.Aiming at the problem that the dark channel prior theory is not applicable to large areas of the sky,an image segmentation algorithm based on fuzzy set theory is proposed,which combines morphological image processing to reasonably segment the foggy image into the sky area and the non-sky area.The transmittance of the sky area is optimized,and the atmospheric light value is estimated combined with the sky area;On the other hand,in view of the problem that the soft matting algorithm takes too long to optimize the rough estimation of transmittance,a guided filtering is used to optimize the rough estimation of transmittance,which ensures the effect of fog removal and improves the operating efficiency of the algorithm.Furthermore,the effectiveness of the algorithm is proved by simulation experiments.4.Aiming at the computer vision system with high real-time requirements,a fast image dehazing algorithm based on Gaussian blur is proposed.The algorithm only uses one Gaussian blur to estimate the transmittance,and then combines with a fast atmospheric light value estimation method to restore the image,which can ensure the dehazing effect and greatly shorten the running time of the algorithm.Finally,the effectiveness of the algorithm is proved by simulation experiments.