Do You Need an Infrared Camera for rPPG?

No, you do not need an infrared camera for most rPPG use cases. If your goal is heart-rate estimation during a telehealth visit or other structured daytime workflow, a normal RGB camera is often enough.

Infrared becomes attractive when the environment gets harder or the measurement ambition gets bigger. Low light, nighttime monitoring, skin-tone robustness, and advanced multispectral use cases can justify it. But a lot of teams jump to infrared before they have exhausted what a good RGB workflow can already do.

Why This Question Keeps Coming Up

rPPG people learn quickly that ordinary webcams are messy. Lighting changes, skin tone matters, compression hurts, and motion ruins otherwise pretty demos. So infrared sounds like the obvious fix.

Sometimes it helps. Sometimes it just adds hardware cost while the real problem stays unsolved.

That is why the better question is not "Is infrared better?" It is better for what, under which conditions, and at what cost?

For background, see green vs red vs infrared PPG, rPPG lighting conditions and accuracy, rPPG multispectral imaging, and camera-based rPPG.

Why RGB Still Dominates Practical rPPG

The short answer is availability.

Every laptop, phone, and tablet already has an RGB camera. That matters for deployment more than many engineers want to admit. A perfect sensor nobody has is often commercially worse than a decent sensor everyone already owns.

The earliest landmark rPPG work used ordinary ambient-light video. Verkruysse, Svaasand, and Nelson showed in 2008 that pulse could be extracted remotely from facial video using ambient light alone (DOI: 10.1364/OE.16.021434). Poh and colleagues pushed the field further with webcam-based heart-rate measurement from RGB facial recordings (DOI: 10.1364/OE.18.010762). De Haan and Jeanne later improved robustness with the CHROM method, which is explicitly built around RGB chrominance behavior (DOI: 10.1364/BOE.4.001781).

That history matters because it shows the field did not begin with fancy infrared rigs. It began with RGB because RGB was available and already good enough to prove the principle.

RGB has another advantage: color information

This part gets missed in hardware conversations. Many successful rPPG methods do not just look for a pulse in one video stream. They use the fact that pulse and motion affect red, green, and blue channels differently. CHROM and related approaches lean on those channel relationships to suppress motion and illumination drift. A single monochrome infrared feed may do better in the dark, but it also removes some of that color-based separation unless you add more bands or different model assumptions.

So infrared is not simply "RGB, but better." It is a different signal design with different tradeoffs.

What RGB Is Good At

Heart rate in decent lighting

This remains the mainstream win. If the patient is still, front-lit, reasonably centered, and captured at enough quality, RGB can recover a strong enough pulse signal for heart-rate estimation.

Low-friction telehealth deployment

If you want a browser-based workflow across everyday consumer devices, RGB is the obvious choice. Nobody wants to ship an external infrared camera just to read pulse before a visit.

Lower total system complexity

RGB-only systems reduce hardware, integration, driver, and device-support headaches. That matters in clinical operations.

Where Infrared Helps

Infrared is not pointless. It just needs to be justified.

1. Low-light and nighttime capture

Near-infrared systems can perform better when visible-light conditions are poor or unstable. That matters for nighttime monitoring, sleep-adjacent workflows, and some vehicle or home environments.

2. Skin-tone robustness

Melanin absorbs visible light more strongly than near-infrared. That is one reason researchers are interested in NIR rPPG for reducing performance disparities. Sun et al. reported that NIR rPPG at 850 nm showed much smaller skin-tone-related error differences than green-channel RGB systems in their study cohort (DOI: 10.1109/TBME.2023.3245897).

This does not mean infrared automatically solves fairness. It means the optical starting point may be better in some settings.

3. Multispectral or research-grade measurement goals

If the goal moves beyond heart rate into perfusion mapping, oxygen-related estimation, or advanced physiology research, multispectral systems become more attractive. Kumar et al. and Shao et al. both reported controlled studies using infrared-oriented camera setups for contactless oxygen-related measurement tasks, results that are interesting but far more specialized than standard webcam telehealth (DOI: 10.1145/2733373.2806332, DOI: 10.1109/TBME.2020.3023951).

What Infrared Does Not Magically Fix

This is where marketing gets annoying.

Motion artifact

If the face moves, the signal still breaks. Infrared does not repeal geometry.

Compression

If your video pipeline crushes subtle temporal information, infrared will not rescue it.

Bad workflow design

If the patient sits with a window behind them, talks during capture, or keeps drifting out of frame, that is a product problem. Buying a different wavelength will not fix that.

Overclaiming on SpO2

People often hear "infrared" and think "great, now it is a pulse oximeter." No. Clinical pulse oximetry depends on controlled red and infrared optics, calibration curves, and validated hardware. A generic infrared camera does not automatically become an oximeter. That is why our answer in can a camera measure oxygen saturation is still mostly no for routine care.

When RGB Is the Right Call

RGB is usually the right answer if:

• you need to run on existing phones, tablets, or laptops
• the primary metric is heart rate
• the measurement happens in a guided daytime workflow
• low deployment friction matters more than marginal optical gains
• you are still proving adoption, completion, and product-market fit

This covers a lot of telehealth and digital intake use cases.

When Infrared Is Worth Paying For

Infrared becomes more compelling if:

• measurements happen in low or uncontrolled visible light
• you need better robustness across a wider range of skin tones and environments
• you are building a dedicated device rather than a browser-first software product
• you care about night monitoring or longer-duration observation
• you are pursuing research-grade multispectral physiology, not just pulse

That is a smaller set of use cases than many vendor decks imply.

What infrared adds besides wavelength

Teams sometimes budget only for the camera module. Real NIR deployment usually means more than that: active illumination, power management, thermal considerations, optics that pass the chosen band, device integration, and software access to the sensor. If you want a web workflow on ordinary laptops and phones, that alone may kill the plan because most consumer browsers expose the standard color camera, not a dedicated NIR stream.

There is also a user-experience tradeoff. A browser-first RGB flow can run on devices patients already own. A dedicated NIR setup often pushes you toward kiosk, bedside, or purpose-built hardware. That can be the right move for sleep labs, inpatient observation, or a regulated device program. It is usually the wrong move if you are still trying to prove completion rates in outpatient telehealth.

A practical way to choose

If your use case is browser telehealth during the day, start with RGB. If it is a clinic kiosk with controlled lighting, test RGB first and add NIR only if failure rates stay high after workflow fixes. If it is overnight monitoring or an embedded device you control, NIR deserves a serious look. If the product claim depends on oxygen saturation, do not let either RGB or generic NIR cameras carry that burden without purpose-built validation.

The Buyer Mistake to Avoid

The classic mistake is buying infrared because the RGB demo looked shaky.

If your RGB demo failed because of poor framing, poor lighting guidance, or weak quality rejection, the real fix is product discipline. Teams often jump to hardware when they should first improve capture instructions, exposure handling, and signal-quality checks.

On the other hand, if your actual deployment environment is dim, nighttime, or inherently variable, then yes, infrared may be the right upgrade.

The order matters.

First fix workflow. Then decide if wavelength is still the bottleneck.

My Recommendation

For most commercial rPPG deployments, especially telehealth and browser-based intake, start with RGB.

Use infrared only when one of three things is true:

1. the environment is too hard for RGB
2. the measurement goal is harder than plain heart rate
3. you control the hardware and can justify the added complexity

That is the sane path.

Infrared is a valuable tool. It is not the default answer to every rPPG problem. A good RGB system beats a badly deployed infrared one. And for many companies, the fastest route to real value is still a well-designed RGB workflow that runs on devices patients already own.

So do you need an infrared camera for rPPG?

Usually no.

Do some high-performance and specialized systems benefit from it? Absolutely. Just do not let a wavelength debate distract you from the harder truth: most failures in rPPG are still workflow, validation, and product honesty problems.

References

1. Verkruysse W, Svaasand LO, Nelson JS. "Remote plethysmographic imaging using ambient light." Optics Express 16(26) (2008). DOI: 10.1364/OE.16.021434
2. Poh MZ, McDuff DJ, Picard RW. "Non-contact, automated cardiac pulse measurements using video imaging and blind source separation." Optics Express 18(10) (2010). DOI: 10.1364/OE.18.010762
3. de Haan G, Jeanne V. "Robust pulse rate from chrominance-based rPPG." Biomedical Optics Express 4(10) (2013). DOI: 10.1364/BOE.4.001781
4. Sun X, et al. "Near-infrared rPPG and skin-tone robustness." IEEE Transactions on Biomedical Engineering (2023). DOI: 10.1109/TBME.2023.3245897
5. Kumar M, et al. "Contactless oxygen saturation estimation with infrared-oriented imaging." Proceedings of UbiComp (2015). DOI: 10.1145/2733373.2806332
6. Shao D, et al. "Dual-channel near-infrared imaging for contactless oxygen-related estimation." IEEE Transactions on Biomedical Engineering (2020). DOI: 10.1109/TBME.2020.3023951