The Research Of Single Image Dehazing Algorithm Based On Boundary Constraints

As an important carrier of information, image is very important for human being and equipment to get information. It is widely used in artificial intelligence, pattern recognition, intelligent transportation and other fields. The quality of an Image directly determines the accuracy of human beings to getting target information. In our life, because of the factors of imaging equipment or bad weather such as haze, fog, images acquired by equipment will appear blurred and degraded. Therefore, it is necessary to study how to eliminate the influence of the haze, so that the foggy image becomes visible and clear.Image dehazing can be done in two ways: one is to enhance the foggy image, the other way is to study the imaging model, and then achieve the purpose of defogging by assumptions or prior knowledge. The former is simple, but the result is not good and with poor scene, the latter is complex but the quality is better, and it can obtain a relatively clear and fog-free image.This thesis analyzes the advantages and disadvantages of these two kinds of methods by introducing. Through the research on image enhancement and image evaluation system, based on the analysis of imaging model, combined with the boundary constraint theory, the thesis focuses on the research of the algorithm based on boundary constraints and context regularization. This work improves the adaptability and the performance of the algorithm. The function of the improved algorithm is realized by experiment and compared with other algorithms. The main research of this thesis include the following aspects.(1) The method of suppressing the noise in the process of image dehazing is studied. The thesis reduces noise by pre-processing with the adaptive median filtering, and using the Laplacian operator to sharpen the foggy image. The experimental results confirm that the noise can be effectively reduced by noise reduction pretreatment of a fog image.(2) The thesis studies the algorithm based on the boundary constraint and context regularization, and solves the problem by improving the algorithm. The proposed algorithm enhances the adaptability of processing different sizes of image by regional adaptive block, and a bilateral filtering is used to optimize the transmission rate based on the boundary constraints in order to keep the image edge structure clear. Experiments shows that the improved algorithm can effectively solve the problem of the original algorithm, and restore the higher quality image.(3) A contrast adjustment method is used to solve the problems such as the image quality degradation and the decrease of the image contrast. The image quality can be improved effectively by the adaptive contrast stretching.(4) This thesis evaluate the image quality after dehazing by different parameters objectively,and compare the improved algorithm with the others from two angles: the quality evaluation of a single image and comparative analysis of multiple images.