全息环纹噪声的空频协同智能抑制
打开文本图片集
关键词：数字全息；环纹噪声；相位成像；傅里叶神经算子；卷积神经网络中图分类号：TH744；O436 文献标识码：Adoi：10.37188/OPE.20253315.2342 CSTR：32169.14.OPE.20253315.2342
Intelligent spatial-frequency cooperative suppression of holographic ring noise
CHEN Benyong，XIONG Zhuang，HUANG Liu， ZHANG Yanchao，FU Xiaping （School of Information Science and Engineering （School of Cyber Science and Technology）， Zhejiang Sci-Tech University，Hangzhou 3lOOl8，China） Correspondingauthor， E -mail：huangliu@zstu.edu.cn；yczhang@zstu.edu.cn
Abstract： In digital holographic imaging，ring noise produced by the difraction of microscopic scaterers is nonlinearly amplified into structured phase errors during numerical reconstruction，thereby limiting the accuracy of quantitative phase imaging and three-dimensional reconstruction. While speckle-noise suppression has been extensively studied，systematic theoretical modeling and targeted suppression strategies for ring noise remain underdeveloped. A novel convolutional neural network，FUResNet，is proposed to operate jointly in the spatialand frequency domains.A multi-scaterer difraction-field superposition model is formulated to accurately simulate ring-noise formation. FUResNet integrates Fourier neural operators，a residual-learning architecture，and atention mechanisms to suppress ring noise eficiently while preserving essential holographic features with high fidelity. Experimental evaluation on simulated and experimental holograms demonstrates that FUResNet significantly outperforms existing approaches：background-noise standard deviation is reduced by 73.9% ，peak signal-to-noise ratio（PSNR） is increased by 13.46dB ， and structural similarity index measure（SSIM） is improved by 13.9% . These improvements across noise suppression，image fidelity，and structural preservation indicate that FUResNet provides an effective solution for high-accuracy quantitative phase imaging.
Key words： digital holography； ring noise；phase imaging； fourier neural operator;convolutional neural network
1引言
数字全息术（DigitalHolography，DH）是一种融合光学干涉记录与计算机数值重建的三维成像技术，可精确提取物光波前的空间振幅分布与相测量位延迟，为三维形貌与折射率测量提供全矢量场数据[1-2]，广泛应用于生物医学成像[3]材料科学4和工业检测[56]等领域。（剩余15554字）
