How rPPG Works in Different Lighting: What Affects Your Scan

Your phone camera picks up something your eyes miss entirely: the faint color shifts in your skin caused by blood pulsing through capillaries with each heartbeat. That's the basis of rPPG, or remote photoplethysmography, and it works remarkably well under the right conditions. But lighting plays a bigger role than most people realize. Whether you're scanning in a bright office, a dimly lit bedroom, or next to a window on a sunny afternoon, the light hitting your face directly shapes the quality of the signal your camera can extract. Understanding how rPPG works in different lighting, and what affects your scan, helps you get more reliable readings every time.

"Most rPPG methods are based on the skin reflectance model, which assumes sufficient ambient lighting to capture pulsating color changes. In low-illumination conditions, signal fidelity decreases significantly." -- Hu et al., IEEE Transactions on Biomedical Engineering, 2024

Why lighting matters so much for rPPG

rPPG depends on detecting tiny variations in how your skin reflects light. When your heart beats, blood rushes into the capillaries near the surface of your face, changing the amount of light absorbed — particularly in the green wavelength range. Between beats, blood recedes and the reflected light shifts again. Your camera records these sub-pixel color fluctuations, and algorithms extract a pulse waveform from them.

The problem is that this signal is small. Really small. The color change between a heartbeat peak and trough might amount to less than 1% variation in pixel intensity. Anything that degrades the camera's ability to capture those tiny shifts — low light, harsh shadows, rapidly changing illumination — makes the algorithm's job harder.

Research by Wang, den Brinker, Stuijk, and de Haan at Eindhoven University of Technology demonstrated this clearly in their 2017 paper in IEEE Transactions on Biomedical Engineering. Their Plane-Orthogonal-to-Skin (POS) algorithm was specifically designed to handle real-world lighting variability by separating the pulse signal from changes caused by ambient light fluctuations. Before POS, earlier methods like ICA-based approaches (Poh, McDuff, and Picard at MIT, 2010) worked well in controlled lab settings but struggled when lighting shifted during a recording.

How different lighting environments affect your readings

Not all light is the same from an rPPG perspective. Here's what the research tells us about specific conditions.

Bright, even indoor lighting

This is where rPPG performs best. Overhead fluorescent or LED lights that illuminate the face uniformly give the camera sensor a strong, consistent signal to work with. A 2022 study published in Applied Sciences by researchers at Bournemouth University (Shervin Yadanpour and colleagues) tested rPPG signal quality across illuminance levels ranging from 5 to 1000 lux. They found that signal morphology remained stable and reliable above roughly 100 lux — the equivalent of a normally lit living room.

Natural daylight near a window

Daylight is bright, but it introduces two complications. First, natural light intensity fluctuates — clouds pass, the sun moves, reflections shift. Second, sitting near a window often means one side of your face is brighter than the other. This asymmetry forces the algorithm to work harder to separate the pulse signal from the gradient. That said, if you're facing the light source rather than sitting sideways to it, daylight generally produces good results.

Dim or low-light environments

Low light is the biggest challenge. When illumination drops below about 50 lux — think a bedroom with a single lamp — the camera's sensor noise starts to overwhelm the already faint pulse signal. The signal-to-noise ratio (SNR) drops sharply. Hu et al. (2024) confirmed this in a study published in IEEE Transactions on Biomedical Engineering, showing that most existing rPPG methods see significant accuracy degradation in low-illumination scenarios, though newer deep-learning approaches are closing the gap.

Mixed or changing lighting

Watching TV in a dark room, walking past windows, or sitting under flickering lights all create rapid illumination changes. These are particularly problematic because they can mimic the frequency of a heartbeat. If ambient light flickers at 1-2 Hz (60-120 changes per second in some lighting), the algorithm might confuse lighting artifacts with actual pulse signals. The CHROM method developed by de Haan and Jeanne (2013, published in Biomedical Optics Express) was one of the first to explicitly address this by using chrominance-based signal combination rather than raw color channels.

Lighting condition	Approximate lux	rPPG signal quality	Key challenge
Bright office (overhead LED/fluorescent)	300-500 lux	Strong	Minimal — best-case scenario
Well-lit living room	100-300 lux	Good	Minor noise, generally reliable
Natural daylight (facing window)	500-2000+ lux	Good to strong	Fluctuations from clouds/movement
Natural daylight (side-lit)	Variable	Moderate	Facial shadows create uneven signal
Dim bedroom (single lamp)	20-50 lux	Weak	Camera noise overwhelms pulse signal
Near-darkness (phone screen only)	<10 lux	Poor	Insufficient reflected light
Flickering or mixed sources	Variable	Variable	Lighting changes mimic pulse frequency

Beyond lighting: other factors that affect your scan

Lighting is the single biggest environmental factor, but it doesn't act alone. Several other variables interact with illumination to influence scan quality.

Skin tone and melanin

Melanin absorbs light, which means darker skin tones reflect less light overall. In low-light conditions, this effect compounds — there's already less light available, and more of it gets absorbed before it can carry pulse information back to the camera. A 2021 study by Nowara, Marks, Mansour, and Veeraraghavan at Rice University (published in npj Digital Medicine) found measurable performance gaps across skin tones when using older rPPG algorithms, though newer methods trained on diverse datasets show significant improvement.

Motion

Moving your head during a scan changes which pixels correspond to which skin regions frame-by-frame. Combined with uneven lighting, motion creates artifacts that look a lot like pulse signals. Staying still matters more in marginal lighting than in bright conditions, where the strong signal can survive some movement.

Camera quality

Not all phone cameras handle low light equally. Newer phones with larger sensors and better computational photography can maintain usable image quality at lower lux levels. The gap between a recent flagship phone and a budget model from a few years ago is substantial when it comes to low-light rPPG performance.

Distance from the camera

Farther away means fewer pixels covering your face, which means less spatial data to extract the signal from. In bright light, this matters less because each pixel has a strong signal. In dim light, you want your face to fill more of the frame.

What the research says about improving low-light rPPG

Researchers have been tackling the lighting problem from multiple angles. Some of the more interesting recent work includes:

Temporal shift attention networks. A 2024 paper by researchers at Sun Yat-sen University (published in IEEE Transactions on Circuits and Systems for Video Technology) introduced neural network architectures that specifically learn to separate lighting variations from physiological signals over time, improving heart rate estimation accuracy by roughly 30% in low-light scenarios compared to traditional methods.

Synthetic data augmentation. Several groups, including work from Philips Research in Eindhoven, have trained rPPG models on synthetically generated video that simulates various lighting conditions. This exposes the model to edge cases it wouldn't see enough of in real-world training data.

Infrared and near-infrared cameras. Some researchers are exploring NIR imaging as a way to bypass visible-light limitations entirely. NIR can detect blood volume changes even in darkness, though consumer phones don't currently have dedicated NIR cameras for this purpose.

Adaptive ROI selection. Rather than using the entire face, newer algorithms dynamically select the skin regions with the best signal quality based on current lighting. If one cheek is in shadow, the algorithm focuses on the well-lit forehead.

Practical tips for better scans

Based on what the research shows, a few straightforward adjustments make a real difference:

• Face your light source. If you're near a window, turn toward it rather than sitting sideways. Even lighting across your face produces cleaner signals.
• Avoid backlit situations. Sitting with a bright window behind you puts your face in shadow from the camera's perspective, even if the room feels bright to you.
• Turn on a lamp if it's dim. Adding one extra light source to bring a room above 100 lux makes a measurable difference in signal quality.
• Hold still for the scan duration. This matters most in moderate or low lighting. In bright, even light, slight movements are less disruptive.
• Keep your phone at arm's length. Close enough that your face fills the frame, far enough that the camera can focus properly.
• Avoid scanning under flickering lights. Older fluorescent tubes flicker at rates that can interfere with pulse detection.

Frequently asked questions

Can I take an rPPG scan in complete darkness?

Practically, no. rPPG relies on reflected visible light, so if there's no light hitting your face, the camera has nothing to work with. Your phone's screen provides some illumination, but it's inconsistent and usually not enough for a reliable reading. A small lamp or overhead light makes a big difference.

Does sunlight work better than artificial light for rPPG?

Not necessarily. Bright, stable artificial light (like LED overhead fixtures) often produces more consistent results because the illumination doesn't fluctuate. Sunlight is bright but variable. If you can sit in consistent indirect daylight, it works well. Direct sun with shifting shadows is less ideal.

Do different phone brands handle low-light rPPG differently?

Yes. Phones with larger camera sensors, wider apertures, and better image processing pipelines capture more light per frame, which gives the rPPG algorithm a stronger signal to work with in marginal conditions. There's no universal ranking since it depends on the specific rPPG software implementation, but hardware quality matters.

Does skin tone affect how much lighting I need?

Research suggests that individuals with darker skin tones may benefit from slightly brighter lighting to compensate for higher melanin absorption. The gap between skin tones narrows substantially in well-lit environments (above 200-300 lux), and modern rPPG algorithms trained on diverse datasets are significantly reducing this disparity.

Where contactless scanning is headed

The lighting problem is one of the most active areas of rPPG research right now. Deep learning models are getting better at separating signal from noise in tough conditions, and some early work with phone-based computational photography — the same technology that makes night-mode photos look good — could eventually be applied to rPPG capture.

Circadify is working on these challenges as part of its contactless vitals platform, bringing rPPG-based scanning to consumer and enterprise applications. For individuals interested in trying camera-based health scanning, the technology is already practical in everyday indoor environments. You can learn more and try it at circadify.com.

If you found this useful, you might also want to read about what happens during a 30-second face scan or how rPPG compares to a pulse oximeter.