Ghost Image
Computational ghost imaging reconstructs an image from known illumination patterns and measurements of the total transmitted or reflected intensity. My work combines structured speckle illumination with deep learning to study reconstruction at low sampling ratios.
0.8% Nyquist computational ghost imaging
I designed and programmed CNN-based computational ghost-imaging workflows using simulated training data. The work investigates reconstruction at 0.8% Nyquist sampling and compares illumination patterns and reconstruction quality under different noise conditions.
Published: Haotian Song et al. “0.8% Nyquist computational ghost imaging via non-experimental deep learning.” Optics Communications 520, 128450 (2022).
Deep correlated speckles
I contributed to research on learned speckle-pattern generation, spatial correlations and reconstruction-quality evaluation. The related publication studies suppression of correlation fluctuations and optical diffraction.
Published: Xiaoyu Nie, Haotian Song et al. “Deep correlated speckles: suppressing correlation fluctuation and optical diffraction.” Photonics Research 12(4), 804–812 (2024).
Related experimental work on speckle disorder is described on my quantum research page.