The Research On Haze Image Features And Image Defogging Algorithm

In recent years, the occurrence of frequency of haze weather increases year by year, it caused serious influence for people’s daily life, haze weather is slowly becoming the city’s meteorological disasters. Under this environment, people take photos in their daily life, the normal work of outdoor surveillance system and some other fields of people take photographs in the fog environment, the reliability of the information collected have being a very serious challenge. The resulting image will be lost a lot of important image feature information because of fog, and affecting the quality of the captured image seriously, at the same time, it has brought great difficulty on the latter part of the image recognition applications, it formed security monitoring risks of human life. Therefore, to carryout an image defogging technology research has very important significance.At this stage, there are two main methods of image defogging. One is based on the enhancement of point of view of image of non-physical model, starting from the image features, the method of enhancing image contrast. Another method is the image restoration which based on a physical model. Through this paper, these two methods are analysed and researched, mainly for the physical model method to make the necessary improvements to the defogging algorithm based on the dark channel prior conditions. This paper mainly includes the following contents:(1) To the existing algorithm to summarize and analyze the fog, pointed out some problems existing in the traditional image defogging algorithm and shortcomings, and make some modification to solve these problems. Based on the existing algorithms are summarized, with dark channel prior to use in this paper the defogging algorithm.(2) The introduction and analysis of atmospheric scattering model, formed by a mathematical model to study the haze image. According to the research of fog image, further analysis of image evaluation mechanism. From the subjective and objective two aspects to obtain, on the image quality with the fog image is analyzed, which can scientifically evaluate.(3) The dark channel prior algorithm available for simulation, and find out the deficiency. Through the analysis and Research on the method of insufficient place, for some improvements on the method. One is in the atmosphere are bright enough light condition or most of the background color and the sky similar circumstances, the original method is not suitable for. The second is the original method proposed soft matting low computational efficiency and high time complexity, which greatly reduces the efficiency of image processing.For the first problem in this paper, the method is setting a tolerance threshold on the bright areas, if the image pixel tolerance is less than the current setup tolerance value, said the current valuation of the transmittance and the actual deviation is big, need to re on the transmission of the estimate. If the pixel tolerance is larger than the current setting tolerance value, said transmission rate estimation is more accurate again without improvement. Aiming at the problems of the second low computation efficiency. In this paper, the using bilateral filtering of the steerable filter to replace the soft matting optimization transmittance.With the help of MATLAB software, the simulation experiment of the improved algorithm. And through the image evaluation mechanism, analysis and comparison of the experimental results obtained before and after. The experimental results show that, the algorithm proposed in this paper, the image defogging effect and processing speed are ideal. Finally combined with faster processing speed, higher efficiency and other characteristics of the VB platform, the improved algorithm in the VB platform operation. It is found that the algorithm on VB platform significantly improve the processing speed, and it is more convenient for the user interface.