Image Defogging Algoritlun Based On Dark Channel And Color Attenuation Prior

Because of its low definition and fuzzy details,fog imaging has a great impact on image analysis and measurement.Therefore,it is of great theoretical significance and practical value to study a fast and effective image defogging algorithm.In this paper,there are two main algorithms of defogging,one is the dark channel prior defogging algorithm,the other is the color attenuation prior defogging algorithm,these two algorithms are better in the field of image defogging and as the most frequent comparison of the two algorithms.In this paper,the principles and defects of the two algorithms are analyzed in depth,and based on this,an improved method is proposed1)A dark channel defogging algorithm based on sky correction and gradient guided filtering is proposed.In order to solve the problem of using fixed value dark channel window in dark channel algorithm,a method of adaptive selection of dark channel window size is proposed.Aiming at the problem of sky distortion caused by dark channel algorithm,super pixel segmentation and Harris corner detection are used to divide the sky area,and a more universal method of sky area transmission correction is proposed.In order to solve the problem of block transmission in non sky area,a new method is proposed to calculate the transmission of non sky area by using average saturation theory and edge detection.In terms of transmittance optimization,a guided filtering method based on gradient information is proposed.According to the texture characteristics of different regions of the image,the appropriate filtering parameters are selected.The optimization results are better than the results of using fixed parameters.At the same time,the guide map is improved to reduce the calculation time.Finally,the transmittance of the image is used as the weight for fusion enhancement.Compared with the original algorithm,the new visible edge ratio and average gradient ratio are increased by about 0.3 and 0.4 respectively.Compared with the algorithm in Chapter 4,the algorithm in this chapter is more suitable for foggy images with more sky regions and scene boundaries.2)A prior defogging algorithm based on color attenuation is proposed in this paper.In order to solve the problem of inaccurate atmospheric light location,we use Otsu algorithm,edge extraction,morphological expansion and image fusion to locate atmospheric light accurately.A method based on fog concentration model is proposed to select the atmospheric scattering coefficient adaptively,which can adaptively select the defogging intensity according to the fog concentration in different regions.This paper improves the time-consuming problem of the original algorithm,and proposes a method of depth of field correction.The curvature filter is used to optimize the depth of field image.Three kinds of curvature filtering are compared.Finally,Gaussian curvature filter is selected for optimization.Considering the shortcomings of the traditional enhancement operator,a multi-scale pyramid fusion algorithm is used and improved.Compared with the original algorithm,the algorithm in this chapter has significant advantages in color,saturation and subjective vision,and the percentage of saturated pixels is about a quarter of that of the original algorithm.Compared with the algorithm in Chapter 3,the algorithm in this chapter is more suitable for foggy images with uneven fog concentration and special highlight region.