The Research Of Image Dehazing Algorithm And Its Application

Photographs taken in the foggy day have the low image contrast and the blurred image content and are gray to some extent. The purpose of the defoggy algorithm is to restore the image contrast and true color under the ideal weather condition. Since lots of algorithms in image processing and computer vision, such as image understanding, object recognition, object tracking, intelligent navigation and so on, only deal with the images and videos captured in the fine day, the image haze removal is very important and is a hot research issue in these fields. In recent years lots of dehazing algorithms are presented continually in the top international journals and conferences in computer vision and image processing.This paper analyzes deeply the physical process of image formation in the foggy weather conditions and reviews the atmospheric scattering image degradation physical model in foggy weather conditions and some conventional image enhancement algorithms, image dehazing algorithms using multiple images and a single iamge. Through deeply study of image dehazing algrithms in the recent ten years we propose an algorithm based on Bayesian framework using sparsity priors to realize single image dehazing.For a foggy input image, there is a clear image correspondingly. The Bayesian framework is established using the maximum probability of the corresponding clear image appearing for a given foggy image. Each probability items in the Bayesian framework have the specific meanings. The clear natural image statistics have sparp peak and long tail characteristics and sparsity priors depict this property properly. The noise of foggy images can be considered simply as a white gaussian noise and local scene depth can be considered smooth. We use MAP method, IRLS algorithm and alternating Iterative optimization method to solve the optimization problem in Bayesian framework.In order to further illustrate the effectiveness of the algorithm, we do many comparison expeiments with three algorithms proposed in the recent international conference and analyze their advantages and disadvantages.Through the comparative analysis, this paper prove the effectiveness of the algorithm furtherly.