基于FPGA的低照度EBCOMS图像噪声处理算法  

Noise Image Processing Algorithm for a Low-light EBCMOS Acquisition System Based on FPGA

作　　者：曹益 朱香平 张笑墨 赵卫 马俊 CAO Yi;ZHU Xiangping;ZHANG Xiaomo;ZHAO Wei;MA Jun(State Key Laboratory of Transient Optics and Photonics,Xi'an Institute of Optics and Precision Mechanics,Chinese Academy of Sciences,Xi'an 710119,China;University of Chinese Academy of Sciences,Beijing 100049,China;Xi'an Institute of Atomic Precision Manufacturing Technology Co.,Ltd.,Xi'an 710110,China)

机构地区：[1]中国科学院大学西安光学精密机械研究所瞬态光学与光子技术国家重点实验室,西安710119 [2]中国科学院大学,北京100049 [3]西安中科原子精密制造科技有限公司,西安710110

出　　处：《光子学报》2024年第11期259-273,共15页Acta Photonica Sinica

基　　金：国家自然科学基金国家重大科研仪器研制项目(No.52127817);中国科学院重大科研仪器设备研制项目(No.ZDKYYQ20220007);陕西省自然科学基础研究计划一般项目(No.2023-JC-QN-0747);陕西省重点研发计划(No.2021LLRH-03);科技重大专项(No.KJZD20231023100501003)。

摘　　要：针对器件研发过程中因材料和工艺技术等不完善导致夜视图像中存在复杂噪声的问题,以及对比传统图像处理的软件平台算法的实时性低、成本高等特性,分析了中国科学院西安光机所自主研发的低照度EBCMOS成像特点,并基于现Xilinx-FPGA设计了一种多阶段脉冲噪声抑制和边缘增强算法,专门针对EBCMOS在低照度条件下采集的图像中存在泊松噪声、椒盐噪声和散粒噪声等混合噪声等问题。实验结果表明,所提算法相比于中值滤波和高斯滤波峰值信噪比分别提高了11.37%和26.64%。与基于高端处理器软件平台相比,该算法处理一帧图像的速度提升了约20倍,可降低成本和实现实时有效处理夜视噪声图像的目的,还可以为实现穿戴夜视设备的集成化和轻量化提供技术支撑。The development of Electron-Bombarded Complementary Metal-Oxide-Semiconductor(EBCMOS)technology marks a significant advancement in image sensors,particularly enhancing low-light and night vision applications.This paper addresses challenges in device development due to material and manufacturing imperfections that introduce complex noise into night vision imagery.Additionally,it highlights the drawbacks of traditional software-based image processing platforms,including their low real-time performance and high operational costs.An analysis of low-light imaging characteristics of EBCMOS,developed by the Xi′an Institute of Optics and Mechanics,is presented.This research introduces a multi-stage pulse noise suppression and edge enhancement algorithm,designed on a Xilinx-FPGA platform,tailored to tackle mixed noises such as Poisson noise,salt-and-pepper noise,and speckle noise prevalent in low-light conditions.EBCMOS sensors are distinctive in their ability to enhance visibility in dark environments through an electron-bombarding mechanism that amplifies signals before CMOS processing.This capability is crucial for applications requiring high-quality night vision,such as security surveillance and wildlife monitoring.The imperfections in materials and fabrication processes can result in various noise types,degrading image quality and obscuring critical details.The paper details a sophisticated FPGA-based algorithm that leverages modern processing power to efficiently reduce noise while preserving important image details.This is achieved through advanced noise reduction techniques that specifically target the unique characteristics of each noise type,improving upon traditional methods like median filtering and Gaussian blurring.Experimental results show that this algorithm enhances the Peak Signal-to-Noise Ratio(PSNR)by 11.37%over median filtering and 26.64%over Gaussian blurring.Moreover,the processing speed for a single image frame is improved approximately 21 times compared to software platforms on high-end processo

关 键 词：低照度EBCMOS 图像降噪 FPGA 中值滤波 高斯滤波 

分 类 号：TN15[电子电信—物理电子学]