Fourier-domain Crosstalk and Leakage Correction for Binary-Encoded Fourier Single-Pixel Imaging
Abstract
Digital micromirror devices enable high-speed Fourier single-pixel imaging (Fourier-SPI), but their binary modulation introduces quantization errors that degrade the reconstructed image quality.
In particular, each binarized Fourier probe contains non-target Fourier components, thereby distorting the corresponding Fourier-domain measurement.
This Letter presents a Fourier domain correction framework that characterizes these errors using the Fourier-domain response matrix of the decoded binary Fourier patterns.
Accordingly, a response-matrix-based Fourier-SPI measurement model is established by partitioning the matrix into components associated with the sampled and unsampled frequencies, thereby decomposing the quantization error into sampled-frequency crosstalk and unsampled-frequency leakage.
This model established a generalized correction framework, termed Fourier Crosstalk and Leakage Correction (FCLC), alongside two simplified methods: Fourier Sampled-Frequency Crosstalk Correction (FSCC) and Fourier DC Crosstalk Correction (FDCC).
FSCC is implemented via an inversion-free Richardson-type iteration, whereas FDCC admits a closed-form solution.
Numerical simulations and experiments across multiple binary encoding strategies reveal that the proposed framework improves reconstruction fidelity without modifying the binary patterns or acquisition process.