Image Dehazing Algorithm Based On Dark Channel Prior And Retinex

Human can obtain all kinds of external information through images very intuitively.Whether the image observed by the human eye is clear or not will have a certain impact on the subsequent various image processing processes.In recent years,with the intensification of global warming,the environment in various regions of China has deteriorated,and fog and haze weather have frequently appeared.Bad weather has caused the performance of many outdoor camera equipments to decline significantly,which will ultimately affect the quality of images obtained by human.In order to enable the outdoor imaging system to work stably in harsh fog and haze weather,and thereby obtain high-quality fog-free images that do not affect the viewing experience,the fog degraded images need to be defogged.The prior knowledge of dark channel can provide theoretical support for image dehazing.Based on the dark channel prior theory and the Retinex enhancement algorithm,two new defogging algorithms are proposed in this paper.The specific research contents are as follows:(1)Aiming at the problem that the traditional defogging method is easy to produce edge blur and poor restoration of details when defogging foggy images,a new algorithm for image defogging based on adaptive hybrid constraint joint variational defogging model and Retinex enhancement algorithm is proposed.Firstly,the weighted fusion method is used to jointly process the two different characteristics of transmittance images from the foreground and sky regions to obtain a rough estimate of the defogging image,then the adaptive hybrid constraint joint variation model is used to make a fine estimate of the transmittance image,and finally the fog-free image is obtained by calculating the atmospheric scattering model.The display effect of the fog-free image is further enhanced by the Retinex algorithm.The proposed alternating direction method can effectively solve the complex optimization problem of the entire defogging process.Experimental results show that the proposed new algorithm can effectively protect the image structure and enhance the display of image detail.Experiments verify that,for artificial fog images and natural fog images,the proposed new defogging algorithm is more competitive than traditional defogging algorithms in preserving image edge details and removing noise interference.(2)Aiming at the defogging method based on dark channel prior theory and traditional variational model is likely to cause staircase effects in the smooth area of the image during defogging,a new image dehazing method based on adaptive total generalized variational regularization and Retinex enhancement algorithm is proposed.The proposed new model directly combines two adaptive total generalized variational regularities in one energy functional,that means transmission images estimation and processes of defogging can be performed simultaneously.The resulting image is further enhanced by the Retinex algorithm to get better display effect.In order to speed up the running efficiency of the program,a primitive-dual algorithm is proposed.Experimental results show that the proposed new method can effectively dehaze natural images,and experimental results verify that the new method can not only make the details of the distant view position of the uniform fog image more obvious than the traditional dehazing method,it can also effectively remove the dense fog in the image containing non-uniform fog.