Research Of Image Haze Removal Improved Algorithm Based On Dark Channel Prior

With the rapid development of Chinese economy, China has achieved a great success and ranked the second largest economy entity in the world. However, environment deterioration and haze come along with the huge success, which badly affects the image collection and management in traffic monitoring. Under this circumstance, it is of great significance and priority to conduct my research.To deal with the restoration of fog degraded image problem, two mainstream methods are adopted: image enhancement and image restoration. Besides, the typical algorithm among numerous way of settling issues is single image haze removal algorithm using dark channel prior. This algorithm enjoys great priviledge for its simpleness in theory, and haze removal efficiency. Nevertheless, the algorithm can not be applied for all images, for example, failure will occur when the image with high luminance is expected to do the fog elimination. In addition, the dark channel haze removal algorithm affect user experience for its complicate in calculation, which requires large scale sparse linear equations analysis, moreover, it will lead to speed decreasing in computing, occupying too much RAM storage, and poor real-time performance.This thesis makes a specific clarification of knowledge for haze removal algorithm using dark channel prior, the imaging model in fog weather and optimizing transmissivity, at the same time, it provide suggestion for optimizing the original algorithm in modifying defects which will decrease the speed of fog degrading and hurdle the effect of fog degrading in the high luminance area. In thesis, guided filtering is recommended to replace the method of soft matting, which fix the problem of massive RAM storage. In the process of calculating transmissivity, the methods of downsampling and interpolation are introduced, which reduce the time of fog degrading, meanwhile, with the combination of gradient operator and the dark colors prior regulation that is more suitable for the dark channel prior requirement, to avoid the distortion of image and strengthen the effectiveness of defogging. To settle the issue of low brightness contrast of image after fog degrading, the method of parabola enhancement is adopted to optimize the visual effect of image after fog degrading.