logologo
HomeProjectsPublicationsMCAMContact Us

Quantized Fourier Ptychography with Binary Images from SPAD Cameras

Xi Yang1
Pavan Chandra Konda1
Shiqi Xu1
Liheng Bian2
Roarke W. Horstmeyer1,3,4

Photonics Research (2021)

1Department of Biomedical Engineering, Duke University, Durham, North Carolina 27708, USA, 2School of Information and Electronics, Beijing Institute of Technology, Beijing 100081, China, 3Department of Physics, Duke University, Durham, North Carolina 27708, USA, 4Department of Electrical and Computer Engineering, Duke University, Durham, North Carolina 27708, USA

Paper
Learned Sensing Network (LSN) Framework
Fig.: (a) Schematic of an FP imaging system. Time-sequential illumination from different angles is provided by an illumination array. Each LED shifts a unique part of the sample’s Fourier spectrum [shown by circles in (b)] into the imaging lens, producing corresponding bandpass-filtered sample images [example images in (d)]. The three LEDs are chosen as examples in (a), (b), (d), and (e) with a common color scheme. (c) Both operating modes of SPAD array cameras can generate binary images. (d) Experimental FP data and reconstructions using 12-bit image data. (e) Quantized FP performed with binary image measurements provides reconstruction image quality closely matching 12-bit data in (d).
Learned Sensing Network (LSN) Framework
Fig.: Example images from a simulation dataset demonstrating four exposure strategies considered in this work based on 32 by 32 LED array. LED positions for selected images (a), (b), (c) are shown in (e). The full high-dynamic-range nature of the dataset can be observed in (1). In (2), uniform exposure method results in most of the image pixel values rounding to zero due to quantization. In (3) and (4), max and informed methods use exposures that are dependent on the sample’s frequency spectrum distribution, leading to low quantization error. (5) Max or informed methods are not sufficient when the extreme case of binary is considered, in which case (6) the median method is chosen.
Learned Sensing Network (LSN) Framework

Experimental results from the SPAD array camera. Comparison between the standard FP on the HDR dataset and the quantized FP on the binarized dataset with the medianexposure strategy is shown for three different samples.

logo
© Copyright 2025 Computational Optics Lab | Web Dev by Laminar Systems