Research On GPU-based Lucky Imaging Algorithm

Under the influence of atmospheric turbulence,the imaging quality and resolution of the ground-based optical telescope are severely restricted.With the rapid development of optoelectronic detection devices in recent decades,the post image processing technology has developed.Lucky imaging technology is a relatively simple post image processing technique,which uses an image-quality evaluation function to evaluate quality of the short exposure image.A few of high quality images are selected,shifted and added to give a much higher resolution image.This paper is based on the theory and classical algorithms of lucky imaging technology.According to this,a lucky imaging solution for the 2.4m telescope at Lijiang Observatory is presented.According to the situation of observation target,we determine the short exposure image and equipment configuration plan needed by lucky imaging technology,mainly including choosing the right camera,exposure time and magnification.The algorithm is based on the classical lucky imaging algorithm to make appropriate simplification,making the algorithm more suitable for the short exposure target image in this paper.Furthermore,a GPU acceleration algorithm is designed to conduct parallel computing during the image selecting period.This GPU acceleration algorithm is based on the existing parallel bubble sorting algorithm,parallel summation and maximum search reduction algorithm,which is suitable for the peak sorting algorithm,image superposition algorithm and image peak search algorithm that is needed in the algorithm flow.Finally,a high-resolution image of binary star with an angular separation of 0.3? is obtained.The experimental results show that images taken by 2m large aperture telescope under the near infrared condition can be used to rebuild a high-resolution image(FWHM?0.2?)by lucky imaging technology.A GPU device can accelerate the lucky imaging algorithm,and improve the efficiency of the whole algorithm.