The Reseach Of Single Image-Defog Processing Algorithm And Software Implement

As the continuously improve the performance of computer, computer vision monitoring equipment widely used in road monitoring, military reconnaissance, remote sensing and other fields. However, in the foggy environment computer vision monitoring equipment to obtain the image or video quality would be serious degradation, and directly effects the image feature extraction, reduced the value of images and led to the image features based on target detection, tracking and follow-up work cann’t be normally work. Therefore, the fog image processing and improving image quality appears particularly necessary. This article based on two views of image enhancement and physical model to search the method of single image defogging processing algorithms.This article mainly from the following aspects to dicuss about the topic:1)In aspect of image enhancement for defog, especially based on kinds of classic enhancement algorithm simulation analysis, such as histogram equalization, homomorphic filtering, retinex etc, and summarizes their advantages, disadvantages and the usable range. In the physical model, the paper focuses on the research of a new method to calculate——the Dark Channel Prior(DCP), and also summarizes its advantages and disadvantages through a series of the analysis of the simulation.2)As the use of histogram equalization of image processing for defog, to enhance image contrast will also cause the loss of image detail. Therefor, to design a kind of solution based on homomorphic filtering and histogram equalization is necessary. First we use a modified Gauss high-pass filter as homomorphic filter function to filter the image and enhance image details, then use the Contrast Limited Adaptive Histogram Equalization(CLAHE)to have more image contrast enhancement. The simulation results show that this scheme is better than the original one.3)The DCP is good at the algorithm for defog, but the Soft matting which is one part of the DCP is so complexity that it will cost too much time to achieve the results, these disadvantages now become the bottleneck of the algorithm in a real-time application. In order to solve the questions above, we design two kinds of scheme based on morphological erode and median filter by dark channel prior for defog. the former for the fog of uniform distributed has good effect, the latter to fog effect with the original algorithm is basically the same, compared to the original algorithm to calculate the time-consuming to be reduced by nearly20times, the white balance algorithm has been used for the scheme which cause the color distortion in color image processing. 4)Software of single image defog based on MFC and Opencv is developed and it also realize the relevant improvement scheme to the defog.