Research On Single Image Dehazing Algorithm

At present,the development of information technology and visual image processing technology are changing with each passing day.Image defogging technology is also widely used in various fields.A clear image of the high quality images can convey more details,can make the computer vision system to get a lot of effective information provide a reference for the human production and living,but in the haze environment collected images,often influenced by environmental factors such as the role of lead to its contrast is not high,color distortion and contains a lot of noise,the human eye vision effect is poor.Therefore,the degradation of acquired image quality often produces interference in the process of human recognition,and the computer vision system is also affected in the later image processing and information extraction.Therefore,image processing technology is needed to improve the quality of images.In this paper,the defogging algorithm is studied.Two improved image fog algorithm based on the dark channel priori principle,based on the boundary constraints and guided filtering,the specific research content is as follows:(1)In order to solve the problem of processing failure of sky region in dark channel priori image defogging algorithm,this paper proposes an improved sky region segmentation algorithm.Firstly,the modified Otsu algorithm is used to segment the edge contour of the sky part,and the noise of the foreground area is eliminated by the corrosion operation of morphology and median filtering.Then,the guide filter is used to refine the binary image to retain more details.The obtained sky region is judged by the discriminant formula.If the image is judged to contain the sky region,the atmospheric light value is estimated in the sky region;otherwise,the original algorithm is used.For the images with sky region,different methods of transmittance calculation are used and the transmittance is smoothed.Finally,the recovery image of fog removal is obtained by atmospheric scattering model.The results of theimproved algorithm are compared with those of other algorithms,and the experimental results show that the improved algorithm is better than other algorithms,and has stronger ability to remove fog.(2)To solve the problem of insufficient parameter adaptability in boundary constraint and context regularization image defogging,the detail and clarity of image need to be improved.In this paper,a defogging algorithm based on boundary constraints and guided filtering is proposed.Firstly,in the process of transmittance optimization,different block sizes are selected adaptively for images of different sizes.Then,the guided filter is used to refine the transmittance image to smooth the image edges and retain more details.Then a tolerance mechanism is introduced to correct the transmittance,so that the transmittance is in line with the actual value of the size,the true restoration of bright color or bright areas.Finally,the image brightness adjustment.In the end,you get better resolution,better contrast,and better visibility.Compared with other algorithms mentioned in other literatures,this algorithm achieves better results.The improved algorithm proposed in this paper can obtain satisfactory defogging images.It achieves better visual effects in the subjective.In the objective analysis index,the algorithm in this paper is compared with the existing algorithm,and the improved algorithm in this paper is improved in several objective evaluation indexes.