Researches On Haze Removal Using Dark Channel Prior

With the rapid development of computer vision and image processing, the outdoor machine vision system is widely used in many fields. But the fog seriously influenced the reliability of the outdoor machine vision system, which has brought great influence to people’s life. Therefore, how to deal with the foggy image fast and efficiently is being paid to much more attentions.This thesis mainly introduces three methods on image dehazing that are based on dark channel prior. The first one is an integration of dark channel prior and guiled filtering. In this method, the transmission map and the atmospheric light are estimated according to the dark channel prior, and then a guide filtering is used to get the restored image.The second is a fast algorithm for restoration of foggy images in HSV space. After converting a haze image from its RGB color space to the HSV color space, the transmission map and the atmospheric light are estimated by the saturation component S instead of the dark channel. The third is a new approach for removing haze from images in HSV. The underlying idea of the third mothed is to restore the true color of each pixel by minimizing the lowest of RGB values per pixel. And the method is accomplished using the HSV color space and the haze image model.A comparative study and quantitative evaluation is carried out with different haze images and underwater images. The results demonstrate that all the methods have a great effect in mist images. But for images with a dense fog, only the third method gets a good effect. Through the experiments of underwater images, it can be seen that the three kinds of methods can be applied to the field of underwater image restoration. Besides that, the second method has an obvious advantage in the execution speed then the other methods.It is discovered that there is no method can meet the requirements for the defog effect and the execution speed at the same time. By comparing the algorithms and results of the three methods, the feasibility of the HSV components and the possible development directions of each method are pointed out in this thesis, which proves that this thesis does have certain practical value.