Research On Improved Dark Channel Prior Image Dehazing Algorithm

In a haze environment,the tiny particles in the air absorb and scatter light,which changes the light's propagation path,so that the image and video images generated by the imaging device show a decrease in contrast and color distortion,which will have a great effect on image processing.Therefore,eliminating or reducing image distortion caused by haze weather can improve the overall performance of related systems,which has strong theoretical value and practical significance.In response to these problems,researchers have proposed a variety of image defogging algorithms including a dark channel prior defogging algorithm.Based on the atmospheric scattering model,this thesis conducted further research on the dark channel prior defogging algorithm,and an improvement method was proposed for the problems existing in the algorithm.The specific research contents are as follows:(1)By analyzing the experimental results of the dark channel prior defogging algorithm,it was found that there are problems such as color distortion and darker images in the dark channel prior defogging result.Aiming at these problems,this thesis proposed an improved dark channel prior image dehazing algorithm.First,calculate the variance of the R,G,and B channels of the pixels in the image to determine the fluctuation range.If the fluctuation is large,the minimum value of the three channels is selected as the dark channel pixel value of the current pixel.If the fluctuation is small,the dark channel value of the current point is determined by the average value of the smallest channel in the surrounding pixels,so as to obtain an accurate dark channel image and atmospheric light value.Second,speed up the calculation of the transmittance image by sampling and interpolation;Finally,by converting the image to the HSI color model,image restoration is performed on the intensity,hue and saturation of the image.(2)In video defogging,this thesis introduced the video sequence defogging algorithm based on the limited contrast histogram equalization,and analyzed that the algorithm has the problems of undesirable defogging effect and high time complexity in the process of video defogging.this thesis proposed a video defogging algorithm based on the segmentation of the sky region by combining the dark channel prior algorithm and the three-frame difference method.The algorithm first used the V component image in the HSV color space to segment the sky and non-sky regions in the image,and used the non-sky regions to accurately estimate the atmospheric light value.Secondly,by calculating the degree of change of the pixels and their neighborhoods in the video sequence image,it is determined whether to update the transmittance image of the current foggy image,thereby speeding up the calculation of the transmittance image of the video.Finally,the defogging model restoration formula is used to perform defogging to obtain a clear video sequence without fog.This thesis performed simulation experiments on fogged images and videos in different scenes.The results show that the improved dark-pass prior image defogging algorithm proposed in this thesis was more obvious than the original algorithm in defogging,the speed is faster,and the dark channel prior is solved the algorithm has problems of color distortion and dark image,and is compared with the histogram defogging algorithm and retinex algorithm to demonstrate the efficiency of the algorithm from both subjective and objective aspects;the video sequence defogging algorithm proposed in this thesis can be effective Eliminates the flicker of video sequences,and has an ideal defogging effect and good robustness,and the average processing speed is above 24 frames per second,which can meet the requirements of real-time defogging.