Research On An Image Dehazing Algorithm Based On Saturation Enhancement

With the rapid development of computer vision technology,the outdoor visual systems have been widely used in human’s life,such as intelligent monitoring systems,automatic driving systems,aviation detection systems,etc.However,these systems are usually susceptible to the haze weather,which causes various problems to the captured images,for example,color cast,information loss and detail unclear.Therefore,it has an urgent demand in real life to research a fast and effective algorithm to process the haze or fog image and eliminate the interference of weather conditions.The traditional image dehazing algorithm usually has some problems,which need to be improved,such as cannot correct color cast in images,poor dehazing ability for distant objects in the images and high algorithm complexity.Based on the questions mentioned above and the haze-degraded model,this paper analyzed the reasons for the phenomenon of blur and color cast of images in detail and expounded the feasible for dehazing through desaturation enhancement.Hence,this paper proposes a dehazing method based on saturation enhancement to solve the above question.The specific research contents of this paper are as follows.Firstly,an atmospheric scattering model for image color correction is established.By analyzing the causes of color bias in the haze-degraded image,we reveal that in addition to the atmospheric light and atmospheric transmittance in different color channels,the scene lighting condition on the surface of the photographed object are also the main reasons for the color cast in haze-degraded images.Therefore,the scene illumination factor is used to revise the atmospheric scattering model in this paper.Then,the color correction of the dehazing image can be achieved by balancing the scene illumination of different color channels in the dehazed image.Secondly,a dehazing algorithm based on saturation enhancement is proposed.By analyzing the relationship between the saturation of the image and the haze density of the scene,this paper finds that there is an inverse proportional relationship between the saturation of the image and the haze density.Therefore,enhancing the saturation of the image can significantly restore the haze-degraded image.Base on this,a method is proposed to enhance saturation by reducing the minimum channel pixel value of the image,and then we can obtain the transmittance map through the saturation enhancement image.After that we use an iterative optimization algorithm based on median filtering to refine the transmittance map to smooth the minute details and maintain the depth of field in the transmittance map.Finally we can get the dehazed image.The actual simulation results show that the spatial frequency and average gradient of the dehazed image have reached 26.11 and 5.83.Compared with He’s.,Kim’s,and Zhu’s algorithms,these two parameters improve 4.56%and 8.81%,respectively,the speed of the algorithm improve 12.73%.these parameters prove that the dehazing algorithm has certain advantages in running speed and visual quality of dehazed images.Thirdly,a real-time video dehazing algorithm is studied.This paper simplifies the saturation-based enhancement dehazing algorithm.Gaussian down sampling of the initial image transmittance is used to reduce the computation of the transmittance optimization algorithm.At the same time,this paper uses OpenMP parallel computing and SSE instruction set to optimize and accelerate the algorithm to achieve the purpose of real-time video dehazing.The optimized algorithm can satisfy the real-time dehazing requirements of 720p video under dual-core processors and the real-time dehazing requirements of 1080p video under quad-core processors.