A Real-time Haze Removal Algorithm Of High Resolution Image

With the development of information technology,high-definition surveillance cameras have been applicable to many occasions and can meet many kinds of requirements.For example,they are used in highway monitoring,urban road monitoring,aerial photography and so on.In recent years,the developing countries such as China often suffered from haze weather or other bad weather.In these extreme weather conditions,the high definition camera is affected by the atmosphere that contains small particles and small droplets.The images captured by the cameras usually lost some important information,that means the high-definition cameras fail to capture clear images.Researchers from various countries have been devoted to the enhancement technology of degraded images and put forward some haze removal methods.However,high-resolution images taken by surveillance cameras have a large amount of data,and the dehazing process usually takes a very long time.The existing image haze removal algorithms can not meet the requirements of real-time in practical applications.Therefore,it is necessary to research on the fast haze removal algorithm for high-resolution images.The traditional panorama stitching technique based on manifold can't retain the contents of particular key areas for videos.The ability to express the motion information of video is poor and the speed is slow.To these problems,an efficient panoramic stitching method with significant region constraint is presented.Based on the manifold video stitching and combined with the image feature points,this method first set significant area to generate the panorama with constraint for key frames.Then the alignment and fusion of the key frames' panoramas is needed to obtain the motion panorama.The CUDA parallel computing is used to speed up.The experimental results showed that the proposed method can not only retain the specific form of specific object at a certain moment,but also can keep moving objects of multiple movement patterns in the panorama.And it can effectively improve the speed of image stitching.Aiming at the problem of "halo artifactf" and sky distortion caused by the dark channel haze removal algorithm,this paper proposes a vision-aware haze removal algorithm based on the sky region,which uses the dark channel to segment the sky region of the image,"Iterative segmentation" for the second subdivision,and finally the sky area visual perception optimization.Experiments show that this algorithm can effectively solve the problem of "halo artifact" and sky distortion,restore the natural and natural color of the sky,and achieve an excellent dehazing result.Finally,this thesis achieved a high-resolution image haze removal system with our proposed algorithms.The system is used to dehaze the input image using the image sharpening process,significantly improving the image contrast and sharpness.At the same time the integration of other classic haze removal algorithm,you can use a variety of haze removal algorithm to compare the experimental results,the main image quality and processing time on the comparison to verify the effectiveness of this haze removal algorithm...