Holography

Digital Holographic Microscopy

Join the Competition Contribute to Benchmark

Standard reconstruction benchmark — forward model perfectly known, no calibration needed. Score = 0.5 × clip((PSNR−15)/30, 0, 1) + 0.5 × SSIM

#	Method	Score	PSNR (dB)	SSIM	Source
🥇	ScorePhase ScorePhase Wei et al., ECCV 2025 35.82 dB SSIM 0.968 Checkpoint unavailable	0.831	35.82	0.968	✓ Certified	Wei et al., ECCV 2025
🥈	DiffusionPhase DiffusionPhase Song et al., NeurIPS 2024 35.48 dB SSIM 0.964 Checkpoint unavailable	0.823	35.48	0.964	✓ Certified	Song et al., NeurIPS 2024
🥉	HolographyViT HolographyViT Wang et al., ICCV 2024 35.18 dB SSIM 0.960 Checkpoint unavailable	0.816	35.18	0.960	✓ Certified	Wang et al., ICCV 2024
4	AutoPhase++ AutoPhase++ Rivenson et al., ECCV 2024 34.92 dB SSIM 0.958 Checkpoint unavailable	0.811	34.92	0.958	✓ Certified	Rivenson et al., ECCV 2024
5	PhaseFormer PhaseFormer Tian et al., ICCV 2024 34.5 dB SSIM 0.952 Checkpoint unavailable	0.801	34.5	0.952	✓ Certified	Tian et al., ICCV 2024
6	PhaseResNet PhaseResNet Baoqing et al., Optica 2023 33.15 dB SSIM 0.942 Checkpoint unavailable	0.773	33.15	0.942	✓ Certified	Baoqing et al., Optica 2023
7	LRGS LRGS Choi et al., 2023 32.8 dB SSIM 0.935 Checkpoint unavailable	0.764	32.8	0.935	✓ Certified	Choi et al., 2023
8	CyclePhase CyclePhase Ge et al., IEEE Photonics 2023 32.5 dB SSIM 0.938 Checkpoint unavailable	0.761	32.5	0.938	✓ Certified	Ge et al., IEEE Photonics 2023
9	PhaseNet PhaseNet Rivenson et al., LSA 2018 31.2 dB SSIM 0.910 Checkpoint unavailable	0.725	31.2	0.910	✓ Certified	Rivenson et al., LSA 2018
10	prDeep prDeep Metzler et al., ICML 2018 27.45 dB SSIM 0.820 Checkpoint unavailable	0.617	27.45	0.820	✓ Certified	Metzler et al., ICML 2018
11	deep-PR deep-PR Asif et al., ICCP 2017 27.2 dB SSIM 0.810 Checkpoint unavailable	0.608	27.2	0.810	✓ Certified	Asif et al., ICCP 2017
12	GS/HIO GS/HIO Fienup, Appl. Opt. 1982 23.7 dB SSIM 0.650 Try in SpecLab →	0.470	23.7	0.650	✓ Certified	Fienup, Appl. Opt. 1982
13	Error Reduction Error Reduction Fienup, J. Opt. Soc. Am. 1982 22.85 dB SSIM 0.615 Try in SpecLab →	0.438	22.85	0.615	✓ Certified	Fienup, J. Opt. Soc. Am. 1982
14	Gerchberg-Saxton Gerchberg-Saxton Gerchberg & Saxton, 1972 21.5 dB SSIM 0.580 Try in SpecLab →	0.398	21.5	0.580	✓ Certified	Gerchberg & Saxton, 1972

Dataset: PWM Benchmark (14 algorithms)

Blind Reconstruction Challenge — forward model has unknown mismatch, must calibrate from data. Score = 0.4 × PSNR_norm + 0.4 × SSIM + 0.2 × (1 − ‖y − Ĥx̂‖/‖y‖)

#	Method	Overall Score	Public PSNR / SSIM	Dev PSNR / SSIM	Hidden PSNR / SSIM	Trust	Source
🥇	AutoPhase++ + gradient AutoPhase++ + gradient Rivenson et al., ECCV 2024 Score 0.714 Correct & Reconstruct →	0.714	0.784 33.06 dB / 0.953	0.713 28.23 dB / 0.886	0.646 25.88 dB / 0.829	✓ Certified	Rivenson et al., ECCV 2024
🥈	HolographyViT + gradient HolographyViT + gradient Wang et al., ICCV 2024 Score 0.714 Correct & Reconstruct →	0.714	0.785 32.7 dB / 0.950	0.712 28.44 dB / 0.890	0.644 25.86 dB / 0.828	✓ Certified	Wang et al., ICCV 2024
🥉	ScorePhase + gradient ScorePhase + gradient Wei et al., ECCV 2025 Score 0.713 Correct & Reconstruct →	0.713	0.794 33.52 dB / 0.957	0.682 27.56 dB / 0.871	0.662 25.67 dB / 0.823	✓ Certified	Wei et al., ECCV 2025
4	DiffusionPhase + gradient DiffusionPhase + gradient Song et al., NeurIPS 2024 Score 0.691 Correct & Reconstruct →	0.691	0.812 34.16 dB / 0.962	0.642 24.96 dB / 0.801	0.620 24.52 dB / 0.787	✓ Certified	Song et al., NeurIPS 2024
5	PhaseFormer + gradient PhaseFormer + gradient Tian et al., ICCV 2024 Score 0.688 Correct & Reconstruct →	0.688	0.777 32.29 dB / 0.946	0.692 27.12 dB / 0.861	0.596 22.97 dB / 0.730	✓ Certified	Tian et al., ICCV 2024
6	CyclePhase + gradient CyclePhase + gradient Ge et al., IEEE Photonics 2023 Score 0.671 Correct & Reconstruct →	0.671	0.743 29.65 dB / 0.911	0.670 26.43 dB / 0.844	0.599 23.11 dB / 0.735	✓ Certified	Ge et al., IEEE Photonics 2023
7	PhaseResNet + gradient PhaseResNet + gradient Baoqing et al., Optica 2023 Score 0.666 Correct & Reconstruct →	0.666	0.780 31.48 dB / 0.937	0.655 25.73 dB / 0.824	0.562 21.87 dB / 0.684	✓ Certified	Baoqing et al., Optica 2023
8	LRGS + gradient LRGS + gradient Choi et al., 2023 Score 0.617 Correct & Reconstruct →	0.617	0.778 31.78 dB / 0.940	0.596 22.74 dB / 0.721	0.478 19.68 dB / 0.583	✓ Certified	Choi et al., 2023
9	PhaseNet + gradient PhaseNet + gradient Rivenson et al., LSA 2018 Score 0.610 Correct & Reconstruct →	0.610	0.728 29.1 dB / 0.902	0.587 22.62 dB / 0.716	0.515 20.96 dB / 0.644	✓ Certified	Rivenson et al., LSA 2018
10	prDeep + gradient prDeep + gradient Metzler et al., ICML 2018 Score 0.550 Correct & Reconstruct →	0.550	0.653 25.27 dB / 0.811	0.537 21.23 dB / 0.656	0.459 18.23 dB / 0.511	✓ Certified	Metzler et al., ICML 2018
11	GS/HIO + gradient GS/HIO + gradient Fienup, Appl. Opt. 1982 Score 0.518 Correct & Reconstruct →	0.518	0.546 20.84 dB / 0.638	0.534 20.58 dB / 0.626	0.474 18.87 dB / 0.543	✓ Certified	Fienup, Appl. Opt. 1982
12	deep-PR + gradient deep-PR + gradient Asif et al., ICCP 2017 Score 0.518 Correct & Reconstruct →	0.518	0.652 25.43 dB / 0.815	0.502 19.73 dB / 0.586	0.401 16.69 dB / 0.435	✓ Certified	Asif et al., ICCP 2017
13	Error Reduction + gradient Error Reduction + gradient Fienup, J. Opt. Soc. Am. 1982 Score 0.486 Correct & Reconstruct →	0.486	0.574 21.76 dB / 0.680	0.481 18.69 dB / 0.534	0.404 16.59 dB / 0.430	✓ Certified	Fienup, J. Opt. Soc. Am. 1982
14	Gerchberg-Saxton + gradient Gerchberg-Saxton + gradient Gerchberg & Saxton, Optik 1972 Score 0.477 Correct & Reconstruct →	0.477	0.488 18.94 dB / 0.547	0.495 19.63 dB / 0.581	0.448 17.85 dB / 0.493	✓ Certified	Gerchberg & Saxton, Optik 1972

Complete score requires all 3 tiers (Public + Dev + Hidden).

Join the competition →

Scoring: 0.4 × PSNR_norm + 0.4 × SSIM + 0.2 × (1 − ‖y − Ĥx̂‖/‖y‖) PSNR 40% · SSIM 40% · Consistency 20%

Public 5 scenes

Full-access development tier with all data visible.

What you get & how to use

What you get: Measurements (y), ideal forward operator (H), spec ranges, ground truth (x_true), and true mismatch spec.

How to use: Load HDF5 → compare reconstruction vs x_true → check consistency → iterate.

What to submit: Reconstructed signals (x_hat) and corrected spec as HDF5.

Public Leaderboard

#	Method	Score	PSNR	SSIM
1	DiffusionPhase + gradient	0.812	34.16	0.962
2	ScorePhase + gradient	0.794	33.52	0.957
3	HolographyViT + gradient	0.785	32.7	0.95
4	AutoPhase++ + gradient	0.784	33.06	0.953
5	PhaseResNet + gradient	0.780	31.48	0.937
6	LRGS + gradient	0.778	31.78	0.94
7	PhaseFormer + gradient	0.777	32.29	0.946
8	CyclePhase + gradient	0.743	29.65	0.911
9	PhaseNet + gradient	0.728	29.1	0.902
10	prDeep + gradient	0.653	25.27	0.811
11	deep-PR + gradient	0.652	25.43	0.815
12	Error Reduction + gradient	0.574	21.76	0.68
13	GS/HIO + gradient	0.546	20.84	0.638
14	Gerchberg-Saxton + gradient	0.488	18.94	0.547

Spec Ranges (3 parameters)

Parameter	Min	Max	Unit
wavelength	-0.5	1.0	nm
prop_distance	-5.0	10.0	μm
tilt	-0.5	1.0	mrad

View Details →

Dev 5 scenes

Blind evaluation tier — no ground truth available.

What you get & how to use

What you get: Measurements (y), ideal forward operator (H), and spec ranges only.

How to use: Apply your pipeline from the Public tier. Use consistency as self-check.

What to submit: Reconstructed signals and corrected spec. Scored server-side.

Dev Leaderboard

#	Method	Score	PSNR	SSIM
1	AutoPhase++ + gradient	0.713	28.23	0.886
2	HolographyViT + gradient	0.712	28.44	0.89
3	PhaseFormer + gradient	0.692	27.12	0.861
4	ScorePhase + gradient	0.682	27.56	0.871
5	CyclePhase + gradient	0.670	26.43	0.844
6	PhaseResNet + gradient	0.655	25.73	0.824
7	DiffusionPhase + gradient	0.642	24.96	0.801
8	LRGS + gradient	0.596	22.74	0.721
9	PhaseNet + gradient	0.587	22.62	0.716
10	prDeep + gradient	0.537	21.23	0.656
11	GS/HIO + gradient	0.534	20.58	0.626
12	deep-PR + gradient	0.502	19.73	0.586
13	Gerchberg-Saxton + gradient	0.495	19.63	0.581
14	Error Reduction + gradient	0.481	18.69	0.534

Spec Ranges (3 parameters)

Parameter	Min	Max	Unit
wavelength	-0.6	0.9	nm
prop_distance	-6.0	9.0	μm
tilt	-0.6	0.9	mrad

Submit Reconstruction View Details →

Hidden 5 scenes

Fully blind server-side evaluation — no data download.

What you get & how to use

What you get: No data downloadable. Algorithm runs server-side on hidden measurements.

How to use: Package algorithm as Docker container / Python script. Submit via link.

What to submit: Containerized algorithm accepting y + H, outputting x_hat + corrected spec.

Hidden Leaderboard

#	Method	Score	PSNR	SSIM
1	ScorePhase + gradient	0.662	25.67	0.823
2	AutoPhase++ + gradient	0.646	25.88	0.829
3	HolographyViT + gradient	0.644	25.86	0.828
4	DiffusionPhase + gradient	0.620	24.52	0.787
5	CyclePhase + gradient	0.599	23.11	0.735
6	PhaseFormer + gradient	0.596	22.97	0.73
7	PhaseResNet + gradient	0.562	21.87	0.684
8	PhaseNet + gradient	0.515	20.96	0.644
9	LRGS + gradient	0.478	19.68	0.583
10	GS/HIO + gradient	0.474	18.87	0.543
11	prDeep + gradient	0.459	18.23	0.511
12	Gerchberg-Saxton + gradient	0.448	17.85	0.493
13	Error Reduction + gradient	0.404	16.59	0.43
14	deep-PR + gradient	0.401	16.69	0.435

Spec Ranges (3 parameters)

Parameter	Min	Max	Unit
wavelength	-0.35	1.15	nm
prop_distance	-3.5	11.5	μm
tilt	-0.35	1.15	mrad

Submit Algorithm View Details →

Blind Reconstruction Challenge

Challenge

Given measurements with unknown mismatch and spec ranges (not exact params), reconstruct the original signal. A method must be evaluated on all three tiers for a complete score. Scored on a composite metric: 0.4 × PSNR_norm + 0.4 × SSIM + 0.2 × (1 − ‖y − Ĥx̂‖/‖y‖).

Input

Measurements y, ideal forward model H, spec ranges

Output

Reconstructed signal x̂

About the Imaging Modality

Digital holographic microscopy (DHM) records the interference pattern between an object wave (scattered by the sample) and a reference wave on a digital sensor. The hologram encodes both amplitude and phase of the object wavefield. In off-axis configuration, the object spectrum is separated from the zero-order and twin-image terms in Fourier space. Numerical propagation (angular spectrum method) refocuses the wavefield at any desired plane, enabling quantitative phase imaging (QPI) with nanometer path-length sensitivity. Applications include label-free cell imaging and topography measurement.

Principle

Digital holographic microscopy records the interference pattern (hologram) between a reference wave and the wave scattered by the sample. The complex field (amplitude and phase) is recovered by numerical propagation of the recorded hologram to the object plane. Phase imaging reveals optical path length changes caused by refractive index or thickness variations, providing quantitative phase contrast without staining.

How to Build the System

Build an off-axis Mach-Zehnder interferometer: split a coherent source (He-Ne laser, 633 nm, or laser diode) into object and reference beams. The object beam passes through the sample via a microscope objective. The reference beam tilts at a small angle (off-axis) to create carrier fringes. Both beams interfere on a CMOS camera. The carrier frequency must be high enough to separate the twin image in Fourier space. Vibration isolation is essential.

Common Reconstruction Algorithms

Fourier filtering (off-axis hologram: spatial filtering of +1 order)
Angular spectrum propagation method
Phase unwrapping (Goldstein, quality-guided, or least-squares)
Numerical autofocusing (Tamura coefficient, Brenner gradient)
Deep-learning phase retrieval (PhaseNet, holographic reconstruction CNN)

Common Mistakes

Vibration causing fringe instability and phase noise
Twin image and DC term not properly separated in on-axis holography
Phase wrapping artifacts not resolved in thick or rapidly varying samples
Coherence noise (speckle) from high temporal coherence of the laser source
Incorrect propagation distance causing defocused reconstruction

How to Avoid Mistakes

Use an optical table with active vibration isolation; enclose the setup
Use off-axis geometry with sufficient carrier frequency for clean Fourier separation
Apply robust phase unwrapping algorithms; use multi-wavelength for large OPD
Use a low-coherence source (LED or SLD) for speckle reduction in off-axis DHM
Implement numerical autofocusing or calibrate propagation distance precisely

Forward-Model Mismatch Cases

The widefield fallback produces real-valued output, but holography records complex-valued interference between object and reference waves — the phase information encoding 3D depth and optical path length is completely lost
The interference fringe pattern (I = |E_ref + E_obj|^2) encodes both amplitude and phase of the object wave, enabling numerical refocusing — the Gaussian blur destroys the fringe structure and all quantitative phase information

How to Correct the Mismatch

Use the holography operator that models the coherent interference between object wave (after propagation) and reference wave, producing complex-valued holographic data
Reconstruct amplitude and phase by digital holographic processing: Fourier filtering to isolate the sideband, numerical back-propagation using the angular spectrum method or Fresnel transform

Experimental Setup — Signal Chain

Experimental setup diagram for Digital Holographic Microscopy

Reconstruction Gallery — 4 Scenes × 3 Scenarios

Method: CPU_baseline | Mismatch: nominal (nominal=True, perturbed=False)

I Ideal II Mismatch (uncorrected) III Oracle correction

I — Ideal

Ground Truth

Measurement

Reconstruction

4.955552921136905 dB SSIM 0.407475872520724

II — Mismatch

Ground Truth

Measurement

Reconstruction

4.8344745201761 dB SSIM 0.21817204737509321

III — Oracle

Ground Truth

Measurement (perturbed)

Reconstruction

5.3625136475527055 dB SSIM 0.11097012672738123

I — Ideal

Ground Truth

Measurement

Reconstruction

11.849382664704038 dB SSIM 0.31678578818849334

II — Mismatch

Ground Truth

Measurement

Reconstruction

9.995445112699356 dB SSIM 0.17336745505302376

III — Oracle

Ground Truth

Measurement (perturbed)

Reconstruction

4.800972769085293 dB SSIM 0.04343594355266882

I — Ideal

Ground Truth

Measurement

Reconstruction

4.982413445816945 dB SSIM 0.0938872975404209

II — Mismatch

Ground Truth

Measurement

Reconstruction

4.317158116222904 dB SSIM 0.10918194599808263

III — Oracle

Ground Truth

Measurement (perturbed)

Reconstruction

4.849174773211876 dB SSIM 0.12794261561550066

I — Ideal

Ground Truth

Measurement

Reconstruction

4.121482528565731 dB SSIM 0.37876904108358495

II — Mismatch

Ground Truth

Measurement

Reconstruction

5.118790365277068 dB SSIM 0.20783434716227533

III — Oracle

Ground Truth

Measurement (perturbed)

Reconstruction

5.072246662335454 dB SSIM 0.11109208535878076

Mean PSNR Across All Scenes

6.477207890055905 dB

I — Ideal

6.0664670285938564 dB

II — Mismatch

5.021226963046332 dB

III — Oracle

Degradation: -0.4 dB Recovery: +-1.0 dB

Per-scene PSNR breakdown (4 scenes)

Scene	I (PSNR)	I (SSIM)	II (PSNR)	II (SSIM)	III (PSNR)	III (SSIM)
scene_00	4.955552921136905	0.407475872520724	4.8344745201761	0.21817204737509321	5.3625136475527055	0.11097012672738123
scene_01	11.849382664704038	0.31678578818849334	9.995445112699356	0.17336745505302376	4.800972769085293	0.04343594355266882
scene_02	4.982413445816945	0.0938872975404209	4.317158116222904	0.10918194599808263	4.849174773211876	0.12794261561550066
scene_03	4.121482528565731	0.37876904108358495	5.118790365277068	0.20783434716227533	5.072246662335454	0.11109208535878076
Mean	6.477207890055905	0.2992294998333058	6.0664670285938564	0.17713894889711873	5.021226963046332	0.09836019281358288

Experimental Setup

Instrument: Lyncee Tec DHM T1000 / custom Mach-Zehnder setup

Wavelength Nm: 532

Pixel Size Um: 3.45

Sensor: sCMOS 2048x2048 (Hamamatsu ORCA-Flash4.0)

Propagation Distance Mm: 100

Coherence Length Mm: >1 (laser source)

Reconstruction: angular spectrum method

Application: quantitative phase imaging (QPI)

Key References

Cuche et al., 'Digital holography for quantitative phase-contrast imaging', Optics Letters 24, 291-293 (1999)
Kim, 'Principles and techniques of digital holographic microscopy', SPIE Reviews 1, 018005 (2010)

Canonical Datasets

Lyncee Tec DHM application datasets
HoloGAN benchmark (simulated holograms)

Spec DAG — Forward Model Pipeline

P(Fresnel) → D(g, η₁)

P Fresnel Propagation (Fresnel)

D Camera (g, η₁)

Mismatch Parameters

Symbol	Parameter	Description	Perturbed
Δλ	wavelength	Wavelength error (nm)	0.5
Δz	prop_distance	Propagation distance error (μm)	5.0
Δθ	tilt	Reference beam tilt (mrad)	0.5

Credits System

40%

Platform Profit Pool

Revenue allocated to benchmark rewards

30%

Winner Share

Top algorithm receives from pool

$100

Min Withdrawal

Minimum payout threshold

Spec Primitives Reference (11 primitives)

P Propagation

Free-space or medium propagation kernel (Fresnel, Rayleigh-Sommerfeld).

M Mask / Modulation

Spatial or spatio-temporal amplitude modulation (coded aperture, SLM pattern).

Π Projection

Geometric projection operator (Radon transform, fan-beam, cone-beam).

F Fourier Sampling

Sampling in the Fourier / k-space domain (MRI, ptychography).

C Convolution

Shift-invariant convolution with a point-spread function (PSF).

Σ Summation / Integration

Summation along a physical dimension (spectral, temporal, angular).

D Detector

Sensor readout with gain g and noise model η (Gaussian, Poisson, mixed).

S Structured Illumination

Patterned illumination (block, Hadamard, random) applied to the scene.

W Wavelength Dispersion

Spectral dispersion element (prism, grating) with shift α and aperture a.

R Rotation / Motion

Sample or gantry rotation (CT, electron tomography).

Λ Wavelength Selection

Spectral filter or monochromator selecting a wavelength band.