Research On Image Haze Removal Algorithm Based On Adaptive Transmission

With the development of computer technology,intelligence has become a symbol of life,and various smart devices are gradually applied to people's daily lives.Computer vision plays an important role in many areas,such as surveillance systems,navigation control,and autonomous driving.However,these outdoor image processing systems are extremely susceptible to weather conditions.In bad weather such as fog and haze,there are a large number of medium particles like dust and water vapor droplets in the air,which causes scattering and refraction during the imaging process of cameras and other equipment,it will reduce the quality of the captured image,thereby reducing color fidelity.Due to the loss of details in hazy images,many computer vision applications that depend on optical images cannot work well.Therefore,dehazing methods are highly needed in computer vision applications and computational photography.In this thesis,the physical model of hazy image degradation is studied.Based on liner transformation and the atmospheric illumination prior,three new image dehazing algorithms are proposed.(1)In order to solve the problems of inadequate transmission estimation in bright region and Halo effect at the edge of depth field in dark channel prior algorithm,an adaptive linear transformation image dehazing algorithm based on Gaussian attenuation is proposed.Firstly,establish a linear transformation model between the minimum channel of hazy images and haze-free images to estimate the rough transmission.Secondly,construct a Gaussian function using the minimum channel of hazy images to adaptively compensate the transmission in bright region,and thereby improve the accuracy of transmission.Then cross-bilateral filter is used to eliminate its texture effects.The experimental results show that the algorithm can effectively improve the color distortion in bright region and eliminate the Halo effect at the edge of depth field.And the restored images have obvious details and appropriate saturation.(2)Aiming at the situation that the algorithm of existing image dehazing can not accurately estimate transmission in white and bright areas,restored image with color distortion and loss of details seriously etc,an adaptive transmission dehazing algorithm based on linear transformation is proposed.First,in the YCbCr space,construct an Anti-S type function to scale the luminance component to reduce the influence of the highlighted pixels.Then,the compressed luminance component is enhanced through a linear transformation model,and the Gaussian function is used to convolve the luminance component to obtain an adaptive control parameter.The adaptive parameter adjusts the restored image quality through the luminance information.Verified by simulation experiments,the proposed algorithm can accurately estimate the transmission,effectively remove image fog and improve color distortion in white areas,and restore more details and edge informations of the image.(3)In order to effectively estimate the depth information of the image in the haze scene,restore clear and haze-free images,an adaptive dehazing algorithm for luminance component positive correlation is proposed.The atmospheric illumination in hazy weather mainly has a great influence on the luminance channel in YCbCr colorspace,but has less impact on the chrominance channel.A linear model is built to describe the positive correlation between the luminance channel of the hazy image and its corresponding depth map.Based on this linear model,its corresponding depth map is estimated.Then,through the depth map construct an exponential function to adaptively obtain atmospheric scattering coefficient,so it can obtain a more accurate transmission.Experimental results show that the algorithm can effectively remove the image haze interference,the saturation of obtained image is appropriate,and it also has certain advantages in the objective evaluation index.