Oscillating Dispersion for Maximal Light-throughput Spectral Imaging

Jiuyun Zhang; Zhan Shi; Linsen Chen; Xun Cao

cs.CVMar 16, 2026

Oscillating Dispersion for Maximal Light-throughput Spectral Imaging

Authors: Jiuyun Zhang, Zhan Shi, Linsen Chen, Xun Cao

Organizations: Nanjing University, Nanjing, China · Key Laboratory of Optoelectronic Devices and Systems with Extreme Performances of MOE, Nanjing University, Nanjing, China

Abstract

Existing computational spectral imaging systems typically rely on coded aperture and beam splitters that block a substantial fraction of incident light, degrading reconstruction quality under light-starved conditions. To address this limitation, we develop the Oscillating Dispersion Imaging Spectrometer (ODIS), which for the first time achieves near-full light throughput by axially translating a disperser between the conjugate image plane and a defocused position, sequentially capturing a panchromatic (PAN) image and a dispersed measurement along a single optical path. We further propose a PAN-guided Dispersion-Aware Deep Unfolding Network (PDAUN) that recovers high-fidelity spectral information from maskless dispersion under PAN structural guidance. Its data-fidelity step derives an FFT-Woodbury preconditioned solver by exploiting the cyclic-convolution property of the ODIS forward model, while a Dispersion-Aware Deformable Convolution module (DADC) corrects sub-pixel spectral misalignment using PAN features. Experiments show state-of-the-art performance on standard benchmarks, and cross-system comparisons confirm that ODIS yields decisive gains under low illumination. High-fidelity reconstruction is validated on a physical prototype.

Figures & tables

Panel of Figure 1 — Figure 1: Depiction of different computational spectral imaging systems. From left to right: coded-aperture systems modulate light with an amplitude mask; spectral-filter-based and diffraction-based systems encode spectral information via wavelength-selective filters or diffractive elements, respectively; ODIS encodes spectra through axial prism translation with a grayscale camera.

Figure 2: Optical layout of ODIS. (a) With the Amici prism at the image-conjugate plane, all wavelengths from a scene point converge to a common sensor location, producing a sharp PAN image. (b) Translating the prism along the optical axis to the defocused plane causes each wavelength to focus at a distinct lateral position, yielding a spectrally dispersed measurement.

Figure 3: Overview of PDAUN. (a) PDAUN architecture with $K$ unfolding stages. (b) Denoiser with three-layer U-shaped structure. (c) FFT-Woodbury preconditioned conjugate-gradient solver. (d) Dispersion-Aware Deformable Convolution (DADC) module.

Panel of Figure 4 — Figure 4: Cross-system simulation HSI reconstruction comparison under low illumination ( $9$ lux). The bottom-left illustrates normalized spectrum curves at the position marked by pink triangle in the PAN image. The right shows reconstruction results at spectral bands 476.5nm, 536.5nm, 584.3nm, 648.1nm. Zoom in for a better view.

Method	Type	Params (M)	GFLOPs	PSNR $\uparrow$	SSIM $\uparrow$	SAM $\downarrow$
SIFDB [ 42 ]	Iterative	–	–	28.04	0.7714	19.45
TSA-Net [ 28 ]	E2E	44.252	135.21	35.88	0.9661	7.89
HDNet [ 17 ]	E2E	2.368	154.79	34.62	0.9593	8.86
MST [ 4 ]	Transformer	3.665	16.74	36.41	0.9697	8.49
CST [ 3 ]	Transformer	1.360	16.16	35.23	0.9631	9.04
SPECAT [ 39 ]	Transformer	0.286	11.45	35.17	0.9592	10.27

Table 3: Quantitative comparison of reconstruction algorithms on the ODIS simulation benchmark. All deep unfolding methods are configured with 5 stages. Best results are in bold and second best are underlined .

Figure 5: Real-world reconstruction on the ODIS prototype. (a) Raw measurements under standard illumination ( $\approx$ 774 lux); spectral curves at the position marked by a red sun are compared with spectrometer references. (b) Reconstructed 21-channel HSI and (c) its pseudo-RGB rendering. (d), (e) Corresponding results under low-light conditions ( $\approx$ 6.7 lux). (f) Photograph of the ODIS prototype.

Baseline	DADC	FFT-WB PCG	PSNR $\uparrow$	SSIM $\uparrow$	SAM $\downarrow$
✓	–	–	36.32	0.9684	8.77
✓	–	✓	36.47	0.9681	8.85
✓	✓	–	36.82	0.9702	8.07
✓	✓	✓	37.01	0.9736	8.09

Table 4: Ablation studies. (a) Break-down ablation on PDAUN’s two core components: DADC (dispersion-aware deformable convolution) and FFT-WB PCG (FFT-Woodbury preconditioned CG). (b) Effect of PCG iteration count.

Explore similar work

Compact Snapshot Spectral Imaging with Calibration-Free Aperture Diffraction
Aug 29, 2026Tao Lv, Quan Yuan, Shiqiao Li +5Hyperspectral Imaging Computational Imaging
Nanjing University, Nanjing, 210023, China
Aperture-aware Dispersion 5-D Light-field Imaging Spectrometer
Jul 6, 2026Chenglong Huang, Tao Lv, Jianing Yang +3Image Reconstruction Hyperspectral Imaging
Nanjing University, Nanjing, 210023, China · Key Laboratory of Optoelectronic Devices and Systems with Extreme Performances of MOE, Nanjing University
Compact Low-Cost Hyperspectral Imaging via Angular-to-Spectral Diversity Conversion
Sep 20, 2026Kazuma Fujiwara, Takuya Funatomi, Kazuya Kitano +2Hyperspectral Imaging Computational Imaging
Nara Institute of Science and Technology, Nara, Japan · Kyoto University, Kyoto, Japan

Aug 29, 2026J/K move · Enter open · S save

Compact Snapshot Spectral Imaging with Calibration-Free Aperture Diffraction

Tao Lv, Quan Yuan, Shiqiao Li, Chenglong Huang+4

Nanjing University, Nanjing, 210023, China

Hyperspectral Imaging Computational Imaging Image Reconstruction Algorithm Unrolling

Open paper AskPDF arXiv

Snapshot Spectral Imaging (SSI) provides high-dimensional temporal-spatial-spectral observation to uncover intrinsic physical characteristics. However, its complex system and repetitive calibration requirements hinder edge applications. Here, we propose a compact, cost-effective, calibration-free SSI method, Aperture Diffraction Imaging Spectrometer (ADIS), which consists only of a diffractive lens with a binary mask and a Bayer-filtered sensor, requiring no additional physical footprint compared to standard RGB cameras. ADIS disperses and multiplexes wavelengths, mapping energy to distinct sensor locations, enabling full-resolution recovery from superpixel-level encodings. ADIS directly leverages theoretically computed PSFs to enable calibration-free spectral reconstruction, while tolerating lens-dependent variations across different optical configurations and bridging the gap between simulation and reality. To achieve SSI by solving a sparsely-constrained inverse problem, we introduce the Orthogonal Diffraction-Aware Unfolding Framework (ODAUF) with Voxel Shift Transformer (VST) for improved orthogonal diffraction perception. Integrating VST into ODAUF forms the efficient Orthogonal Diffraction-Aware Unfolding Voxel Shift Transformer (ODAUVST), delivering excellent recovery and reduced parameters. By elaborating on theory, systematic and comprehensive comparing, and demonstrating real SSI results, we validate the superiority of ADIS, achieving calibration-free full-resolution SSI within a commercial camera footprint.

Abstract

Explore similar work

Compact Snapshot Spectral Imaging with Calibration-Free Aperture Diffraction

Aperture-aware Dispersion 5-D Light-field Imaging Spectrometer

Compact Low-Cost Hyperspectral Imaging via Angular-to-Spectral Diversity Conversion