ELECTRO-OPTICS  LABORATORY

Interferenceless coded aperture correlation holography (I-COACH), which utilizes spatial incoherent light illumination to achieve interferenceless and non-scanning three-dimensional (3D) imaging, has revolutionized the field of incoherent holography. However, existing correlation reconstruction algorithms often suffer from significant inter-plane crosstalk from multiple cross-sections of the sample, resulting in poor quality in 3D imaging. Here, we proposed a single-shot 3D reconstruction method of I-COACH via a Wiener deconvolution network combining total variation kernel (TV-K) functional constraint (WienerNet3D/TV-K). The method first performs multiple pairing of Wiener deconvolution and a convolutional neural network (CNN) and then uses the point spread hologram (PSH) library of the I-COACH system and object hologram (OH) as inputs, updating throughout the entire training process to learn the optimal filter and noise regularization parameters. Moreover, a TV-K function constraint is introduced to achieve more high-frequency details of the sample. Both simulation analysis and experimental results demonstrate that the proposed method has excellent performance in suppressing inter-plane crosstalk and significantly improves the signal-to-noise ratio in I-COACH 3D reconstruction. Importantly, this WienerNet3D/TV-K method will provide a useful strategy for the application of I-COACH dynamic 3D imaging in fluorescence microscopy, astronomy and other fields.