Research On Image Enhancement Algorithm In Foggy Environment

Vision is the most important information source of human, with the development of computer hardware and software technology, the importance of image technology in people’s lives rapidly rising. Image processing technology, such as intelligent navigation, face recognition, and target tracking, can bring great convenience to people. However, the development of industrialization and urbanization, making the haze gradually become the norm in people’s lives. Haze not only affect people’s health, will also affect vision. Images acquired by imaging equipment in the fog generally exhibit the condition of low contrast, low color saturation and blur scene information, affecting people’s access to information through images. The prevalence of haze reduces the stability of outdoor image equipment, such as road monitoring, vehicle navigation, and limits the scope of its application. Defogging technology can remove the impact of haze, restore degraded images due to haze, improve the contrast and saturation of the image, so that the image becomes clear again, and has broad application prospects.Among various de-hazing methods, technology of single image de-hazing has the least requirement for imaging devices and external conditions, so that has a broader application space. However, single image dehazing does not have any other input information, and has a higher demand for the algorithm. So in recent years, it has become a major research focus in the field of image processing. This paper summarizes the current status of de-hazing technology at home and abroad, does a depth study on dehazing technology principles, analyses the degradation model of foggy image, provides a theoretical basis for the study of model-based dehazing methods. In this paper, we implement a variety of single image de-hazing methods which have effect significant and is most representative currently, and analyze their characteristics respectively.On the basis of prevlous work, we propose a new dehazing algorithm based on dark channel and incident light assumption. The algorithm scans for atmospheric light area through progressive windows, and can accurately estimate the atmospheric light. We improved the accuracy of dark channel prior by weakening it, and conduct the initial dehazing based on the new assumption, so that the transmission rate will be stretched to the [0,1] range and the transmission estimation problem is simplified. Through the observation and analysis of the foggy images, we proposed the incident light uniform assumption. We estimate the transmission with the assumption by using the illumination estimation algorithm of Retinex. We finally remove the fog with initial dehazing image and transmission image. Experiment shows that this method can deal with foggy images effectively and stably, the restore image has a clear vision and high contrast. In addition, the method has a high operating efficiency, to meet the requirements of real-time applications.Finally, we do analysis and classification on a large number of foggy images. We do experiment on the images with a variety of algorithms, analyze the performance of different algorithms in different application situation, summarize the characteristics of these algorithms. Also, we examine the computational efficiency of these methods with images of different size.