Streak Image Denosing And Reconstruction Using Adaptive Gaussian Guided Filter

3D laser imaging based on streak tube camera has a high usability value in the fields of military or civil application, especially in 3D non-scanning laser imaging. As core device, the picoseconds streak tube camera have characteristics of high time resolution, short response time and high sensitivity. We use picoseconds camera as core device to build the reconstruction system, and the work principle of picoseconds camera as same as streak tube camera, set of streak tube laser imaging system has been established in our lab to obtain raw streak image data. In this paper, we analyzed the principle of streak tube camera and imaging characters. Furthermore, we research the denoising and enhance algorithms of raw streak image as well as data extracting algorithms of streak imagesFirstly, we analyzed the principle of streak tube cameras and imaging characters of picoseconds camera in the paper. Since picoseconds camera is build basis on single slit streak tube and CCD camera, we can use the streak tube working principle to guide and help us to set up the system. Then we analyzed the impact of noise on streak image and reconstruction result, and designed suitable denoising method for streak image.In this paper, we propose a novel denoising algorithm for streak image obtained from picoseconds streak tube cameras. The source of noise of streak images are complex and various, the ordinary algorithms in frequency domain are not suited to streak image and are not satisfied with the demands on data extraction of streak image. We expect that we could restrain noise, keep the details and enhance the boundary of streak image, as well as low time complexity. We analyzed and compared several different styles of edge-preserving algorithms such as bilateral filter, adaptive bilateral filter, guided filter, Gaussian filter, domain transform filter, block matching and 3D filtering(BM3D), non-local mean filter and so on. Then, we propose an adaptive Gaussian guided filter. This filter could greatly remove the noises, as well as enhance the edges and details of the streak images. This algorithm using a recursive implementation of the Gaussian filters in order to reducing the time complexity, and we have been verified that this algorithm is an O(N) time algorithm.Additional, we put forward the image segmentation and data extraction algorithms for denoised streak image. Using the transverse and longitudinal offset of streak image, we could extract intense and distant information, calculating the actual distance of objects and reconstructing the intense and distance images.