光子集成干涉成像图像重构:熵先验  

Image Reconstruction of Photonics Integrated Interference Imaging:Entropy Prior

作　　者：陈天宝 曾雪锋[1] 白莹莹 田明森 张峰[1] 张学军[1] Chen Tianbao;Zeng Xuefeng;Bai Yingying;Tian Mingsen;Zhang Feng;Zhang Xuejun(Key Laboratory of Optical System Advanced Manufacturing Technology,Changchun Institute of Optics,Fine Mechanics and Physics,Chinese Academy of Sciences,Changchun,Jilin 130033,China;University of Chinese Academy of Sciences,Beijing 100049,China)

机构地区：[1]中国科学院长春光学精密机械与物理研究所中国科学院光学系统先进制造技术重点实验室,吉林长春130033 [2]中国科学院大学,北京100049

出　　处：《光学学报》2021年第23期98-104,共7页Acta Optica Sinica

基　　金：国家自然科学基金(61975201);中科院长春光机所国际合作项目。

摘　　要：光子集成干涉成像方法是近年来发展起来的一种以高质量成像和系统扁平化为目标的远场成像方法,有望在同等分辨率下系统的能耗、体积和质量均减小至1/10~1/100。但是现有系统对于高频信号的采样具有稀疏特性,当使用傅里叶逆变换(IFT)求解物体强度分布时,复原的观测目标在锐边处出现Gibbs振铃伪影,从而影响成像质量。为了抑制伪影,提出熵先验并研究熵惩罚特性,利用熵先验并结合光子集成干涉成像的特点设计最大熵方法。为了验证方法的性能,采用性能较好的多层分级孔径排布结构进行仿真模拟,使用峰值信噪比(PSNR)、结构相似性系数(SSIM)和均方误差(MSE)作为像质评价手段。仿真结果表明,最大熵方法可以消除由高频稀疏采样带来的伪影,对于受振铃影响明显的图像,MSE和SSIM有50%以上的提升,PSNR提升10%以上。Photonics integrated interference imaging method is a far-field imaging method developed in recent years,which aims at high quality imaging and system flattening.It is expected to reduce the energy consumption,volume,and weight of the system can be reduced to 1/10~1/100 under the same resolution.However,the existing systems have sparse characteristics for high-frequency signal sampling.When inverse Fourier transform(IFT)is used to solve the object intensity distribution,Gibbs ringing artifact appears in the sharp edge of the restored observation target,thus affecting the image quality.In order to suppress artifacts,entropy prior is proposed and the entropy penalty characteristics are studied.The maximum entropy algorithm is designed by using entropy prior and combining with the characteristics of photonics integrated interference imaging.In order to verify the performance of the method,a multi-layer hierarchical aperture arrangement structure with better performance is used for simulation,and peak signal-to-noise ratio(PSNR),structural similarity coefficient(SSIM),and mean square error(MSE)are used as image quality evaluation methods.Simulation results show that the maximum entropy algorithm can eliminate the artifacts caused by high-frequency sparse sampling.For the images obviously affected by ringing,MSE and SSIM can be improved by more than 50%,and PSNR can be improved by more than 10%.

关 键 词：成像系统 干涉成像 光子集成干涉成像 熵先验 光子集成芯片 最大熵方法 

分 类 号：O436.1[机械工程—光学工程]