Under-Display Cameras: The Slow Disappearance of the Smartphone Notch

01 The Notch Was a Compromise, Not a Destination

The smartphone front face has been shrinking its furniture for fifteen years. Bezels collapsed through the mid-2010s, and the cameras, earpieces, and sensors that lived in the bezel had to go somewhere. Apple's answer in 2017 was the notch - a cutout in the display itself. Android makers tried everything around it: slim notches, teardrops, motorized pop-up modules, flip-up cameras, and finally the static punch hole, a small circular cutout that became the default answer on Android and, later, a pill-shaped island on iPhones.

Every one of those designs shares one assumption: the camera needs a physically unobstructed window, so the display must make a hole for it. The under-display camera (UDC) rejects the assumption. It puts the sensor behind the display, letting the panel cover the lens completely, and then tries to repair the optical damage the panel causes. That is the whole technology: a display that is asked to be partly transparent, an image sensor that is asked to see through it, and a software layer asked to clean up the mess.

It is the last compromise available. Once the camera lives under the screen, there is nothing left on the front face but glass - which is why the industry keeps coming back to the idea even after a decade of mediocre results.

02 Why OLED Makes It Possible at All

An under-display camera is only conceivable because modern phones use OLED, and Wikipedia's technical definition explains why: an OLED is an LED whose emissive layer is an organic compound film between two electrodes, at least one of which is transparent, and whose individual pixels emit their own light with no backlight. There is no thick backlight stack to see through - just a thin film stack, a TFT backplane, and the cathode and anode layers above and below it.

Every one of those layers is at least partially opaque, and the most opaque is the layer most visible to the eye: the cathode, a thin metal film, plus the TFT wiring, which for years made the UDC region show as a faint grid or smudge on otherwise clean screens. The first production implementations, in 2020, were honest about the trade - ZTE and Xiaomi shipped phones where the camera zone was visibly dimmer, coarser, or both, and reviewers measured a full screen brightness difference you could see with the naked eye.

The engineering work since has been a re-patterning exercise: shrink the cathode's coverage in the camera zone, thin it out, and re-route TFT traces around the region, so more light reaches the sensor while the pixel lattice remains dense enough that the eye does not resolve a texture difference. It is a genuine optimization problem, because the two objectives fight each other directly.

03 The Light-Loss Problem: What the Pixels Take

The core physical trade is simple and brutal. A standard OLED pixel area is roughly 40 to 60 percent opaque across its subpixel structures, cathode, and wiring. Even after re-patterning, UDC implementations have historically transmitted only a few percent to perhaps 20 percent of incoming light to the sensor - a loss of one to two full stops of light-gathering compared to an unobstructed lens. A front camera that would be usable in a dim restaurant through a punch hole becomes, under the display, starved.

Starved light has two consequences. First, signal-to-noise: fewer photons means a noisier image, and no algorithm fully recovers information that never reached the silicon. Second, diffraction and haze: light passes through a semi-regular lattice of pixel structures and gaps, which acts like a diffraction grating and scatters the image. The result is the characteristic UDC signature - soft, hazy, low-contrast selfies with a faint rainbow fringing on high-contrast edges - that every reviewer, including the source video below, has demonstrated on production hardware.

04 Computational Compensation: Fixing What the Screen Broke

Because the optics cannot be fully repaired, the industry repairs the photograph instead. Every UDC phone ships a dedicated image pipeline for the front camera, and the pipeline is where most of the product's quality actually lives. The standard sequence runs roughly: a flare and haze correction pass to remove the scattered-light veil; a deblurring network trained to undo the diffraction softening; a demosaic-style restoration that reconstructs detail lost through the pixel lattice; and conventional multi-frame stacking to claw back signal-to-noise ratio, since stacking many short exposures trades the sensor nothing but time.

The catch is that these networks are trained on data from exactly one optical configuration, and they generalize poorly. Move the camera too far from its trained subject distance, point it at an edge case - a bright window behind a face, fast motion, very low light - and the reconstruction produces artifacts no unobstructed camera would make: smeared hair, plastic skin, edges that shimmer. Reviewers, Marques Brownlee most prominently, have spent years demonstrating exactly this gap on production hardware: the concept works, the marketing shots look fine, and then you take a normal photo in a normal room and the invisible camera is the worst camera on the device.

There is a deeper limit, too. Computational restoration is inference, not recovery. The panel destroyed information before the sensor ever sampled it, and the network's job is to hallucinate a plausible image consistent with what survived. That is acceptable for social-media-resolution selfies. It is not acceptable for anything downstream that expects photographic evidence - face-unlock biometrics, for instance, which is precisely why face-authentication systems still want dedicated, unobstructed sensors or structured-light hardware.

05 Why Adoption Lagged: The Economics of the Last Millimeter

If the technology works at all, why did it take so long to leave the lab? Because UDC is expensive at every layer simultaneously. The display side needs a custom panel region - reduced cathode coverage, rerouted traces, and a higher pixel density or different subpixel arrangement to mask the visible seam - which means a separate manufacturing flow for a strip of screen a few millimeters across. The sensor side needs a module tuned for very low light throughput. The software side needs a trained restoration pipeline per device generation. And the buyer-side payoff is purely cosmetic: the same selfie quality minus a visible dot.

That cost-benefit ratio explains the adoption curve. ZTE shipped first in 2020, Xiaomi followed, Samsung put a UDC in its third-generation foldable (where the inner selfie camera could not be a notch or punch hole without marring the main display), and successive generations have appeared mainly in China-first flagships and foldables - the two segments where a full-screen interior display justifies the premium. Meanwhile the mass market standardized on the punch hole, which costs almost nothing and photographs well. The chart below sketches the front-face design lineage and where UDC sits in it.

06 Where the Notch Actually Stands in 2026

The punch hole has not died; it has become the floor. What has changed is the upper tier. Foldables, the fastest-growing premium form factor, have the strongest structural reason to want a UDC - the inner display is the phone's best screen, and a punch hole in it is a permanent flaw in the product's signature feature - so several 2024-2026 foldables ship UDCs on the interior panel. High-end slab flagships increasingly offer UDC variants, and each generation narrows the visible-seam problem: the camera zone's pixel density penalty shrinks, and the brightness differential between the UDC region and the rest of the panel edges toward imperceptible.

It is a slow death for the notch rather than a fast one, for the economic reasons in section 05. The punch hole is cheap, mature, and photographically superior; the UDC is expensive and photographically compromised but aesthetically perfect. In a market where the front face is the product, aesthetics keep winning just enough design bids to keep the technology funded. That is how bezels died, over years, and it is how the notch is dying now - not by disruption but by attrition from the top of the market downward.

07 Limits and What to Watch

The UDC will not fully replace the punch hole until one of two things happens. Either the panel side gets genuinely transparent - new cathode materials, micro-lens structures carved into the substrate to route light around the pixels, transparent-electrode research of the kind presented at display-technical society conferences - or the sensor side gets so sensitive, via larger sensor areas and stacked architectures, that a five-percent light budget is enough for clean output. Neither milestone has a firm date.

Until then, treat marketing claims with a specific skepticism. "Invisible camera" is true of the screen and silent about the photograph. The two honest tests are unchanged since 2020: take a selfie against a bright window, and look at the camera zone of the display at full brightness off-angle. When a production phone passes both without apology, the notch's long goodbye is over. In 2026, some flagships are close; none are there yet.