Research On Sharpening Method Of Haze Image Under Atmospheric Scattering Model

In recent years,with the rapid development of industrial economy,haze phenomenon has become more and more serious and common,especially in autumn and winter.In such a foggy environment,due to the presence of suspended particles in the air,the details of the target image are often blurred and distorted,which brings great difficulties to the determination of the observed target.Then,how to reduce or eliminate the impact of haze on the target image,and how to clear the target image has become very important.However,the cost of simply improving the hardware facilities is very high,and it is more practical to continuously optimize the image sharpness reduction method.Therefore,in order to effectively improve the image sharpness of the degraded image,the authenticity reduction of the target image needs to be realized.In this paper,fog image is divided into four categories: uniform light fog image,uniform dense fog image,non-uniform or synthetic fog image,large sky area or white scene interference image.On the basis of the atmospheric scattering model and the dark channel fog removal method,the ill-condition equation of the four terms and three unknowns of the atmospheric scattering model is solved.That is to say,the atmospheric light value A and transmittance t(x,y)are solved for obtaining known fog-free image on the basis of image processing technology.The main research and innovation points of this paper are as follows:1.Aiming at the problems of fuzzy image details and halo effect in uniform light fog image,a fast bilinear least squares method for fog removal was proposed.Firstly,the spatial LOG edge detection method is used to obtain the approximate range of atmospheric light value A,and the binary tree algorithm is used to obtain the A value.Secondly,the bilinear least square filtering method is used to optimize the initial transmittance,and the atmospheric scattering model is used to restore the defogging effect,so as to complete the comparison and analysis of subjective and objective experiments.The experimental results show that the obtained clear image can restore the complete image foreground details,filter the image noise and weak the halo effect.Besides,the improved method is efficient.2.To solve the problems of noise amplification and detail loss in the image of uniform fog,an adaptive wiener filtering method based on atmospheric light characteristics was proposed.Firstly,the general range of atmospheric light value A is obtained by Gaussian low-pass filtering method,and the value of A is obtained by quad-tree method.Secondly,the adaptive wiener filter method is used to optimize the initial transmittance,and the image morphology method is used to further optimize and enhance the image details to remove the block effect and image noise.The atmospheric scattering model is used to restore the defogging effect.The experimental results show that the obtained clear image not only achieves good results in brightness,denoising degree,edge holding degree and so on,but also gets effective suppression of block effect and image noise.3.To solve the problem of different image blurring degree in non-uniform foggy images or composite foggy images,a defogging method based on multi-features bidirectional deep convolution network is proposed.Firstly,the accurate atmospheric light value A was obtained by using the gray threshold quartering method on the basis of skyline precise search,and then the algorithm self-learning was completed by extracting the features of fog images.And the nonlinear mapping operation is adopted to achieve the reconstruction of transmittance.Secondly,the atmospheric scattering model is used to obtain the defogging effect.The experimental results show that the transmittance obtained is close to the real scene,and the obtained defogging effect is more real,vivid and natural,and it has good stability.4.In view of the special scene of foggy image sharpening,that is,the interference factor of large sky area or white scene in foggy image scene makes the restored image prone to color deviation,color spot and overexposure,etc.An anisotropic Gaussian filtering method on the basis of accurate searching is proposed to defogging.Firstly,the accurate atmospheric light value A was obtained by adopting the gray threshold quad-tree method based on skyline search.In addition,the anisotropic Gaussian filtering method is used to optimize the transmittance,which can improve the sharpness of image details and the oversaturation effect.The atmospheric scattering model is used to restore the defogging effect.Then,the tone mapping method is adopted to adjust the overall effect of the defogging results.Experimental results show that our method can improve the brightness,contrast and sharpness of the results.By analyzing the characteristics of the four foggy image scenes,different methods were studied to achieve the haze removing processing,so that the restored image would be more real,vivid and natural.