Research On Image Haze Removal Algorithm Based On Dark Channel Prior

Image processing technology is increasingly used in industrial,medical,education,transportation and other industries,it brings a lot of convenience to people's lives and work.However,in recent years,the smog weather in China has become more and more frequent,and the images captured by the outdoor collection system have low contrast,color distortion,and sharpness,which brings great problems to the image processing system.Therefore,it is especially important to research the image haze removal technology under smog weather and ensure the stable and reliable work of the computer vision system.After decades of continuous research by scholars at home and abroad,image haze removal technology has achieved good results.The paper takes image haze removl as the main line of research.The defogging algorithm can be divided into two categories.One is the image enhancement algorithm based on non-physical model,which improves the image quality by improving the contrast of degraded images.The other is an image restoration algorithm based on a physical model.Based on the essence of smog image degradation,a mathematical model is established to achieve image clarity.By analyzing the formation mechanism of the haze image and the characteristics of the haze image,the atmospheric scattering model in the foggy weather is studied to prepare for the in-depth study of haze removal.The global histogram equalization algorithm and the adaptive histogram equalization algorithm for limiting contrast are studied deeply.The advantages and disadvantages of the histogram equalization algorithm in haze removal are verified by experimental simulation.The principle analysis and simulation experiments of the single-scale and multi-scale Retinex algorithm are carried out to verify the dehazing effect of the Retinex algorithm.For the image haze removal algorithm based on the dark channel prior,the atmospheric light intensity is easily interfered by the white object in the image.Soft matting is used to optimize the rough transmittance,but it has low efficiency.The color of the restored image is dark.An improved algorithm is proposed in the paper,based on the dark channel prior theory,combined with the atmospheric scattering physical model,the quad-tree search algorithm is used to accurately estimate the atmospheric light,and the guided filtering algorithm is used to refine the transmission.Finally,a clear and defogged image is obtained.Experiments show that the improved algorithm not only can effectively remove the haze part of the image,restore the essential information of the image,but also enhance the edge details of the image,which is more in line with the visual experience of the human eye.In addition,the real-time performance is also significantly improved.