基于归一化的快速高信噪比关联成像方法研究

The object image can be reconstructed by measuring the intensity correlation function between a bucket detector and a reference detector. Compared with the traditional imaging ways, ghost imaging has many specific characteristics. It can realize super-resolution imaging, and has obvious advantages in anti-transmission medium interference of atmospheric turbulence and scattering. And also it can achieve the lensless imaging. So ghost imaging has potential applications in military detection, medical imaging, microscopic imaging, precision detection, security and so on. However, ghost imaging needs a mass of sample data and massive calculation, but also has a low signal to noise ratio (SNR). Therefore, this project studies three methods to improve the performance of ghost imaging. Firstly, we propose a method, which can be called the positive–negative correspondence normalized ghost imaging, to reduce the storage space and increase the utilization rate of image information. Secondly, normalized iterative denoising ghost imaging is built to effectively improve the quality of the reconstructed image. Thirdly, in order to reduce the ghost imaging time, we use the normalized and fast ghost imaging based on parallel computing. It is expected that the SNR of the existing differential correlation imaging can be improved two to four times, meanwhile the imaging speed can also be raised two to four times comparing with differential ghost imaging. These corresponding results will promote ghost imaging technology in practical processes.

关联成像利用光场的二阶关联性质，将光场的涨落信息与目标物体光强的总探测值关联计算即可成像。关联成像不仅可以实现超衍射极限成像，而且有抗大气湍流、散射等传输介质干扰的优势，也可实现无透镜成像。所以，关联成像在军事探测、医学影像、显微成像、精密检测、安防等领域有着潜在的应用价值及前景。但是热光关联成像需要大量的采样数据及运算，而且成像信噪比较低。因此，本项目研究三种提高关联成像性能的方法：（1）设计基于正负对应的归一化关联成像基本方法，不仅减少关联成像计算所需的数据量，而且有效提高图像信噪比；（2）建立基于迭代的归一化去噪关联成像模型，进一步提高图像质量；（3）利用基于并行计算的归一化快速关联成像思想，将关联成像时间进一步减少。预计本项目可将现有差分关联成像的信噪比提高二到四倍，成像速度提高二到四倍。相应成果推动关联成像技术实用化进程。

关联成像利用光场的二阶关联性质，将光场的涨落信息与目标物体光强的总探测值关联计算即可成像。关联成像不仅可以实现超衍射极限成像，而且有抗大气湍流、散射等传输介质干扰的优势，也可实现无透镜成像。但是在现有的关联成像方法中，归一化关联成像通过计算物体透过率的绝对值，将物臂光路中每一次的桶探测值相对参考臂光路的探测值作归一化，这样可以有效去除光源功率浮动所带来的噪声。相比传统关联成像的效果，归一化关联成像方法所重构的图像信噪比高有数量级的提高。但是归一化关联成像质量依赖于物体的透过率相对方差和桶探测器的信号，并且对光场的涨落很敏感，时变噪声来自于光源强度或者是空间光调制器的效率。同时，该方法重构时间仍然无法满足工程中的实时应用。因此，本项目在深入研究归一化关联成像方法的理论基础上，先设计基于正负对应的归一化关联成像基本方法，不仅减少了关联成像计算所需的数据量，而且有效提高了图像信噪比；然后建立了基于迭代的归一化去噪关联成像模型，进一步提高了图像质量；同时，提出基于并行计算的归一化快速关联成像思想，将关联成像时间进一步减少；最后搭建了基于空间光调制器的关联成像系统，通过数值仿真和实验验证，将现有差分关联成像的信噪比和速度明显提高。本项目对快速高信噪比关联成像方法研究，不仅提高了关联成像的质量，并减少了其图像重构的时间，进而提升成像灵敏度、距离和精度等关键技术指标，推动了关联成像工程化应用，为促进国防科技的创新，满足我国科技工业的建设、以及国民经济建设的长远发展具有重要意义。与此同时，对快速高信噪比关联成像新方法研究，直接促进关联成像技术在生命科学、航空航天科学等相关学科的发展，提高成像探测任务的整体收益，也将为它在雷达精确制导、遥感、医学、显微成像，以及弹载、星载成像探测等相关领域的应用奠定理论基础。