Research On Single Image Dehazing Algorithm Based On Gaussian Weight Decay

With the rapid development of information technology,intelligence has become a symbol of life.Computer vision plays an important role in video surveillance,driverless,image reconstruction and other fields.In actual situations,the outdoor monitoring system is susceptible to haze interference,resulting in images captured by the outdoor monitoring system that does not conform to the human visual perception,such as reduced detail information,loss of features,contrast and color degradation,which can seriously affect the normal operation of the monitoring system.It even leads to paralysis of the monitoring system,which has a great impact on people's life.Thus,in order to make the outdoor monitoring system have certain effectiveness and reliability under hazy weather conditions,it is necessary to dehaze the degraded images captured under hazy conditions.At present,for the research of single image dehazing algorithm,the dark channel prior algorithm is one of the classical algorithms.Through statistical analysis,this method proposes a prior condition to simplify complex problems and achieves effective results,but the algorithm still has certain defects.This paper studies based on the dark channel prior theory and proposes two improved algorithms.?1?In order to solve the Halo effect phenomenon in the depth field of the dark channel prior dehazing algorithm,an adaptive Gaussian attenuation image dehazing algorithm based on edge preservation is proposed.This method introduces the luminance of the sky.Firstly,the accurately edge region of the luminance of the sky is obtained by the edge detection operator,the edge region is no longer processed,for the non-edge region,the adaptive Gaussian function is constructed by the spatial proximity between adjacent pixels to smooth the attenuation.And then the edge regions and the non-edge regions are combined to get the final refined luminance of the sky.It can be seen from the comparative analysis of the experimental results that the proposed algorithm has achieved certain advantages in subjective evaluation and objective evaluation respectively.?2?Aiming at the block effect and low estimation of the transmission in bright regions brought by the dark channel prior dehazing algorithm,an adaptive compensation image dehazing algorithm based on Gaussian weight attenuation is proposed.Firstly,use the edge detection operator to obtain the edge information;Then the Gaussian weight attenuation function is constructed by the edge information to approximate the dark channel;Secondly,using the ratio relationship between the approximation result and the Gaussian function,the bright region is estimated by adaptive compensation of multi-scale luminance components,thereby the coarse compensated estimate transmission is obtained.Finally,the weighted function of the L₁ regularization constraint is used to optimize the coarse transmission to obtain the refined transmission,and combined with the atmospheric scattering model to recover a clear haze-free image.The experimental results show that this proposed algorithm can eliminate the block effect and improve the color distortion problem in bright areas,and the restored image has obvious details information and appropriate brightness.