Research On An Improve Algorithm For Image Dehazing Based On Dark Channel Prior

With the continuous development of network technology and social economy,people's requirements for image definition are increasing constantly.The interference of hazy weather on object imaging results in the decrease of image definition.The image dehazing algorithm is an important method to improve image definition.It has become one of the research hotspots in the field of image processing.Image dehazing refers to the process of improving the contrast and color saturation of images by performing certain processing on foggy images.It can be widely used in mobile phones,cameras,outdoor video surveillance,real-time traffic systems,satellite imaging,etc.At present,a mature image dehazing algorithm is an image dehazing algorithm based on the theoretical model of atmospheric scattering.In this thesis,the incident light attenuation model and atmospheric imaging model are analyzed,and the specific process of of image dehazing is illustrated.Three typical image dehazing algorithms are studied: image dehazing algorithm based on atmospheric dissipation coefficient,image dehazing algorithm based on loss function and image dehazing algorithm based on dark channel prior.Through detailed research analysis and comparative analysis,we find that the image dehazing algorithm based on dark channel prior has better defogging effect.The atmospheric light estimation of the image dehazing algorithm based on dark channel prior falls on the white object or the sky region,so that the atmospheric light is inaccurate and affects the dehazing effect.Therefore,the horizontal and vertical gray projection methods are used to estimate the atmospheric light value.First,the input image is horizontally projected,and sum the horizontal projection.Next,the largest area is selected for vertical projection,and sum the vertical projection value.Then,the largest area is selected and the pixel values are arranged in descending order.Finally,the average gray value of the first 0.1% pixels with the largest brightness value is chosen as the atmospheric light value.The experimental results show that this method can obtain the atmospheric light value closer to the actual situation.For the problem that the dark channel prior law is not satisfied in bright areas such as sky region and the estimation transmittance by dark channel prior is smaller than the actual value in the bright areas,tolerance threshold is used to determine whether the pixel meets the dark channel prior.The theory of dark channel prior is used to estimate the transmittance of pixels in the non-sky region;and the transmittance of the point that does not satisfy the dark channel prior theory is amplified and corrected to make it closer to the actual value for better dehazing effect.The experimental results show that the improved image dehazing algorithm based on dark channel prior has better defogging effect with enhanced contrast and rich detail information.