Research On Image Defogging Method Based On Optimized Atmospheric Scattering Model

With the stable development of China’s economy,the improvement of residents’ consumption ability,and the increasing number of motor vehicles,dust and haze incidents have become increasingly frequent.Image defogging technology is currently an important research topic in the fields of image processing and computer vision.Its main applications include video surveillance,terrain survey,autonomous driving,and object tracking.Haze can cause significant interference to the normal operation of intelligent devices such as machine vision systems,safety detection,and video surveillance,making it difficult to operate stably and efficiently like in fog-free environments.For example,most of the images presented by the traffic control department’s monitoring system under haze conditions have a very significant contrast with the images during clear weather,such as color deviation,overall dimness,and other degradation situations.The model and license plate number of vehicles in the monitoring screen,as well as the clothing and physical features of pedestrians on the road,cannot be accurately recognized,greatly increasing the difficulty of detecting conditions required for facial recognition and action detection in the future,resulting in inaccurate results and causing great trouble for a series of subsequent inspections.The haze environment not only causes serious interference to land transportation,but also has similar impacts on other modes of transportation such as aviation and maritime transportation.In this context,the restoration of degraded images under haze weather is particularly important.The traditional image defogging technology in the foggy environment,when obtaining the foggy image,is based on the calculation of the scene transmission coefficient and the atmospheric light value,assuming that the atmospheric light is constant,and the scene transmission coefficient is dependent on the scene depth.Therefore,compared with the light obtained on the imaging system,the noise light in dust and suspended particles also widely exists in the actual light,and in the actual imaging process,These imperceptible noise lights are often ignored,which enlarges the noise light part and has a great impact on people’s subjective vision.In order to solve this problem,the first thing to do is to optimize the traditional atmospheric scattering model,and add the noise light scattered by the dust and other media in the air that is easily ignored in the past to the traditional atmospheric scattering model,so as to build a new atmospheric scattering model.Next,based on the prior theory of dark primary color,a more advanced solution is created to make the estimation of scene transmittance more accurate.Then,the value of the fog concentration evaluation map of the haze image is segmented by the fog perception density evaluator provided by LIVE data set,and then smoothed by recursive filtering to obtain the fog weight in the air light.The dark channel of the fog layer can be calculated by using the dark channel of the image and the fog weight in the air light.Next,the calculated minimum atmospheric light value in the air can be substituted into the improved atmospheric scattering model,and a more refined scene transmittance can be obtained.Finally,a new objective function is constructed based on the optimized atmospheric scattering model.The objective function is optimized by the total variation denoising model,and a clear fog-free image is obtained.By comparing the evaluation index with other advanced defogging algorithms,it can be verified that the improved algorithm adopted in this paper can not only achieve the purpose of effective defogging,but also suppress the noise existing in the image,and can also play an excellent role in protecting the details of the image,so that the edge information can be well preserved,and has better defogging performance than other defogging algorithms.