Explainer
How phone cameras work: sensors and computational photography
A digital camera sensor turns light into electric signals through the photoelectric effect, then reads those signals out from millions of pixels.[1][2] Most sensors see color through a mosaic of red, green and blue filters, two green to each red and blue, and software fills in the missing colors.[3][4] Phone sensors have small pixels, which limits how much light and contrast they can capture.[5] Phones make up for it with computational photography: they take a burst of short exposures, align them and merge them to cut noise and blur.[6]
This explainer describes how digital image sensors and phone camera software work, based on the makers’ and researchers’ own technical descriptions. It is not a buying guide, and we did not test any camera. Figures for specific sensors and phones are the makers’ claims.
From light to numbers
A camera sensor is a grid of tiny light-sensitive spots called pixels. When light hits a pixel it is turned into an electric signal, a process called the photoelectric effect.[1] The hard part is collecting and reading out those signals from millions of pixels quickly.[2]
The first successful digital imaging sensor was the charge-coupled device, or CCD, invented in 1969 by Willard Boyle and George Smith, who shared in the 2009 Nobel Prize in Physics for it.[7]
Image sensors rely on the photoelectric effect to convert incoming photons to charge.[1] The Nobel committee framed the design problem as collecting and reading out the signals of a large number of pixels in a short time.[2] Boyle and Smith’s 1969 CCD was the first successful solution and earned half of the 2009 physics prize.[7]
In a stacked CMOS sensor a pixel chip of back-illuminated pixels sits on top of a logic chip that holds the signal-processing circuits.[8] In December 2021 Sony announced a design that also splits the photodiodes and the pixel transistors onto separate layers, which it said roughly doubles saturation signal level, widening dynamic range and reducing noise.[9]
How a sensor sees color
On their own, pixels measure only how much light arrives, not its color. So most sensors put a tiny colored filter over each pixel. In the common Bayer pattern, every 2x2 block has two green filters, one red and one blue, because our eyes are most sensitive to green.[3]
Each pixel therefore records only one color. Software fills in the other two by looking at neighboring pixels, a step called demosaicing. Google’s researchers point out that this means two-thirds of a digital photo’s color data is reconstructed.[4]
The layout comes from Bryce E. Bayer’s Kodak patent, US 3,971,065 (1976), in which green-sensitive elements, treated as luminance samples, occupy every other position and alternate with red and blue.[10] The 2x2 repeating unit is two green, one red, one blue.[3] Demosaicing interpolates the two missing channels at each site from nearby pixels, so two-thirds of an RGB image is estimated rather than measured.[4]
Why small sensors struggle
Phone sensors are tiny, and so are their pixels. Google’s camera researchers say this limits dynamic range, the span between the brightest part of a scene that doesn’t turn pure white and the darkest part that doesn’t turn black.[5]
In dim light, phone photos tend to come out blurry or speckled with noise, random variations in brightness from pixel to pixel. A longer exposure collects more light and reduces noise, but it is hard to hold a phone still that long, so the shot blurs.[11]
Google cites the small pixels of smartphone sensors as a limit on their dynamic range.[5] Low light forces a trade-off: noise falls as exposure lengthens, but handheld shake turns long exposures into blur.[11] Sensor makers attack the problem in hardware. Samsung says its 200MP ISOCELL HP2 merges neighboring pixels as light dims, so it can run as a 50MP sensor with 1.2µm pixels or a 12.5MP sensor with 2.4µm pixels.[12] Its Smart-ISO Pro merges a high-ISO and a low-ISO readout taken from a single exposure.[13]
Computational photography: many frames, one photo
Instead of one long exposure, a phone can take a quick burst of short ones. Google’s HDR+, which Google described in 2014 as a camera app feature on its Nexus 5 and Nexus 6 phones, lines the frames up in software and averages them. Averaging cuts noise, and the short exposures keep blur down.[6] Keeping every exposure short also stops bright highlights from turning white, while merging enough frames cleans up the shadows.[14]
Google’s camera starts capturing frames as soon as you open the app, keeping the most recent ones in memory. When you press the shutter, it hands the last 9 or 15 frames to its merging software.[15]
HDR+ captures a burst of short exposures, aligns them algorithmically and replaces each pixel with the average across frames.[6] Rather than bracketing, which blends frames shot at different exposure times, it keeps every exposure short to protect highlights and relies on frame count to clean up shadows.[14] Frames stream into a circular buffer while the app is open, and the shutter press selects the most recent 9 or 15.[15]
Night Sight (2018) extended the method. Google said single-shot phone cameras begin to struggle at 30 lux, and set Night Sight’s target at 3 to 0.3 lux, with no flash.[16] It varied the capture from 15 frames of 1/15 second or less to 6 frames of 1 second, using “motion metering” to measure scene motion and pick exposures that limit blur.[17] Because auto white balance fails in low light, Google trained a learning-based white-balance model.[18]
Zoom without a bigger lens
Ordinary digital zoom just crops the picture and enlarges it, which tends to look blurry or plastic.[19] Google’s Super Res Zoom on the Pixel 3 used the tiny shake of your hand: each frame in a burst lands a few pixels off, and merging them fills in real detail. Google said the result was roughly as good as the 2x optical zoom on many other phones.[20]
Phone makers also add more cameras. Apple’s iPhone 17 Pro, announced in September 2025, has three 48MP rear cameras, with 4x optical zoom at 100mm and an 8x optical-quality setting at 200mm.[21]
Single-frame digital zoom upsamples a crop by interpolation, which smears texture.[19] Super Res Zoom merged a handheld burst directly onto a finer grid, using natural hand tremor, which shifts each frame by a few pixels; Google said the detail was roughly competitive with 2x optical zoom lenses on many other phones.[20] Multi-camera arrays cover longer focal lengths: the iPhone 17 Pro pairs three 48MP cameras with 4x (100mm) optical zoom and an 8x (200mm) optical-quality zoom setting.[21] Apple says its updated Photonic Engine applies machine learning across more of the image pipeline to preserve detail, reduce noise and improve color, especially in low light.[22] For the chips that run these pipelines, see how smartphone chips work.
For more on phones, PCs and the wider device market, see the consumer tech crash course.
Read next
AI PCs explained: Copilot+ PCs, NPUs and the 40 TOPS bar
What "AI PC" and Copilot+ PC mean, Microsoft's 40+ TOPS NPU bar, the Qualcomm, Intel, AMD and Apple chips, Recall, and the debate over usefulness.
Questions readers ask
How does a camera sensor capture light?
Image sensors use the photoelectric effect, which turns light into electric signals. The sensor then has to gather and read out those signals from a very large number of pixels in a short time.[1][2]
Why does a Bayer filter have twice as many green pixels?
In a Bayer mosaic each 2x2 group of pixels has two green, one red and one blue, because human eyes are more sensitive to green. Bryce Bayer's 1976 Kodak patent used the green elements to detect luminance.[3][10]
Why are phone photos noisy in the dark?
In dim light each pixel collects little signal, so photos show random pixel-to-pixel brightness variations called noise. Longer exposures reduce noise but blur handheld shots. Google says the small pixels in phone sensors also limit dynamic range.[11][5]
What is computational photography?
It means using software to combine several captures into one picture. Google's HDR+ takes a burst of short exposures, aligns them and averages each pixel, which reduces noise while short exposures keep blur down.[6]
What is pixel binning?
Pixel binning merges neighboring pixels into one larger pixel when light is low. Samsung says its 200MP ISOCELL HP2 can act as a 50MP sensor with 1.2µm pixels or a 12.5MP sensor with 2.4µm pixels.[12]
Sources
· 22 claims from 6 publishers
Each numbered claim is a statement we checked against the sources listed with it. Tap a number in the text to see its sources in place.
- [1]
Digital image sensors rely on the photoelectric effect, by which light is turned into electric signals. confirmedas of 2009-10-06
- The Nobel Prize in Physics 2009 - Press release · Nobel Prize Outreach · Primary source · 2009-10-06 · Press release (retrieved 2026-11-08)
- [2]
The Nobel committee described the central challenge of an image sensor as gathering and reading out the signals from a large number of pixels in a short time. confirmedas of 2009-10-06
- The Nobel Prize in Physics 2009 - Press release · Nobel Prize Outreach · Primary source · 2009-10-06 · Press release (retrieved 2026-11-08)
- [3]
In a Bayer color filter mosaic, each 2x2 group of pixels has two green pixels, one red and one blue, because human eyes are more sensitive to green. confirmedas of 2018-10-15
- See Better and Further with Super Res Zoom on the Pixel 3 · Google Research · Primary source · 2018-10-15 · Color Filter Arrays and Demosaicing (retrieved 2026-11-08)
- [4]
Google's researchers note that demosaicing fills in the missing colors at each pixel by interpolating from neighbors, so two-thirds of an RGB digital picture is a reconstruction. confirmedas of 2018-10-15
- See Better and Further with Super Res Zoom on the Pixel 3 · Google Research · Primary source · 2018-10-15 · Color Filter Arrays and Demosaicing (retrieved 2026-11-08)
- [5]
Google's researchers say smartphone sensors have small pixels, which limits dynamic range, the span between the brightest highlight that does not blow out and the darkest shadow that does not turn black. confirmedas of 2014-10-27
- HDR+: Low Light and High Dynamic Range photography in the Google Camera App · Google Research · Primary source · 2014-10-27 · Capturing high dynamic range scenes (retrieved 2026-11-08)
- HDR+: Low Light and High Dynamic Range photography in the Google Camera App · Google Research · Primary source · 2014-10-27 · Capturing high dynamic range scenes (retrieved 2026-11-08)
- [6]
Google's HDR+, which Google described in 2014 as a Google Camera app feature for the Nexus 5 and Nexus 6, captures a burst of short-exposure shots, aligns them in software and averages each pixel across the shots; averaging reduces noise and short exposures reduce blur. confirmedas of 2014-10-27
- HDR+: Low Light and High Dynamic Range photography in the Google Camera App · Google Research · Primary source · 2014-10-27 · Introduction (retrieved 2026-11-08)
- HDR+: Low Light and High Dynamic Range photography in the Google Camera App · Google Research · Primary source · 2014-10-27 · Low light section (retrieved 2026-11-08)
- HDR+: Low Light and High Dynamic Range photography in the Google Camera App · Google Research · Primary source · 2014-10-27 · Low light section (retrieved 2026-11-08)
- [7]
Willard Boyle and George Smith invented the charge-coupled device (CCD) in 1969, the first successful imaging technology using a digital sensor, and shared half of the 2009 Nobel Prize in Physics for it. confirmedas of 2009-10-06
- The Nobel Prize in Physics 2009 - Press release · Nobel Prize Outreach · Primary source · 2009-10-06 · Press release, second part of the award (retrieved 2026-11-08)
- The Nobel Prize in Physics 2009 - Press release · Nobel Prize Outreach · Primary source · 2009-10-06 · Press release (retrieved 2026-11-08)
- [8]
A stacked CMOS image sensor places a pixel chip of back-illuminated pixels on top of a logic chip that holds the signal-processing circuits. confirmedas of 2021-12-16
- Sony Develops World's First Stacked CMOS Image Sensor Technology with 2-Layer Transistor Pixel · Sony Semiconductor Solutions · Primary source · 2021-12-16 · News release body (retrieved 2026-11-08)
- [9]
In December 2021 Sony announced a stacked CMOS sensor design that puts photodiodes and pixel transistors on separate layers, which it said roughly doubles saturation signal level, widening dynamic range and reducing noise. confirmedas of 2021-12-16
- Sony Develops World's First Stacked CMOS Image Sensor Technology with 2-Layer Transistor Pixel · Sony Semiconductor Solutions · Primary source · 2021-12-16 · News release body (retrieved 2026-11-08)
- Sony Develops World's First Stacked CMOS Image Sensor Technology with 2-Layer Transistor Pixel · Sony Semiconductor Solutions · Primary source · 2021-12-16 · News release body (retrieved 2026-11-08)
- [10]
Bryce E. Bayer of Eastman Kodak patented a single-sensor color imaging array (US 3,971,065, granted July 20, 1976) in which green-sensitive elements, used to detect luminance, occupy every other position, alternating with red and blue elements. confirmedas of 1976-07-20
- US3971065A: Color imaging array · Google Patents (US Patent and Trademark Office record) · Primary source · 1976-07-20 · Description of the preferred implementation (retrieved 2026-11-08)
- US3971065A: Color imaging array · Google Patents (US Patent and Trademark Office record) · Primary source · 1976-07-20 · Patent record (inventor Bryce E. Bayer; granted 1976-07-20) (retrieved 2026-11-08)
- [11]
Smartphone photos of dim scenes often come out blurry or full of image noise, random pixel-to-pixel variations in brightness, and long exposures that reduce noise tend to blur handheld shots. confirmedas of 2014-10-27
- HDR+: Low Light and High Dynamic Range photography in the Google Camera App · Google Research · Primary source · 2014-10-27 · Introduction (retrieved 2026-11-08)
- HDR+: Low Light and High Dynamic Range photography in the Google Camera App · Google Research · Primary source · 2014-10-27 · Low light section (retrieved 2026-11-08)
- [12]
Samsung says its 200-megapixel ISOCELL HP2 phone sensor automatically merges neighboring pixels as light dims, operating as a 50MP sensor with 1.2µm pixels or a 12.5MP sensor with 2.4µm pixels. confirmedas of 2026-11-08
- ISOCELL HP2 · Samsung Semiconductor · Primary source · Tetra²pixel Technology (retrieved 2026-11-08)
- [13]
Samsung says the ISOCELL HP2's Smart-ISO Pro merges two readouts from a single exposure, combining high-ISO and low-ISO readouts to keep detail in both bright and dark areas. confirmedas of 2026-11-08
- ISOCELL HP2 · Samsung Semiconductor · Primary source · Smart-ISO Pro (retrieved 2026-11-08)
- ISOCELL HP2 · Samsung Semiconductor · Primary source · Smart-ISO Pro (retrieved 2026-11-08)
- [14]
Google says HDR+ uses deliberately short exposures to avoid blown-out highlights and merges enough frames to clean up shadows, instead of bracketing, which blends shots taken at different exposure times. confirmedas of 2014-10-27
- HDR+: Low Light and High Dynamic Range photography in the Google Camera App · Google Research · Primary source · 2014-10-27 · Capturing high dynamic range scenes (retrieved 2026-11-08)
- HDR+: Low Light and High Dynamic Range photography in the Google Camera App · Google Research · Primary source · 2014-10-27 · Capturing high dynamic range scenes (retrieved 2026-11-08)
- [15]
Google described its camera app continuously capturing frames into a circular buffer from the moment it opens, and sending the most recent 9 or 15 frames to its merging software when the shutter is pressed. confirmedas of 2018-11-14
- Night Sight: Seeing in the Dark on Pixel Phones · Google Research · Primary source · 2018-11-14 · Capture pipeline (retrieved 2026-11-08)
- Night Sight: Seeing in the Dark on Pixel Phones · Google Research · Primary source · 2018-11-14 · Capture pipeline (retrieved 2026-11-08)
- [16]
Google wrote in 2018 that single-shot phone cameras begin to struggle at 30 lux and that Night Sight targeted scenes between 3 and 0.3 lux, with a single shutter press and no flash. confirmedas of 2018-11-14
- Night Sight: Seeing in the Dark on Pixel Phones · Google Research · Primary source · 2018-11-14 · What we mean by dark (retrieved 2026-11-08)
- Night Sight: Seeing in the Dark on Pixel Phones · Google Research · Primary source · 2018-11-14 · What we mean by dark (retrieved 2026-11-08)
- [17]
Depending on motion and stability, Night Sight captured anything from 15 frames of 1/15 second or less to 6 frames of 1 second each, and Google's "motion metering" measured scene motion to choose exposure times that limit blur. confirmedas of 2018-11-14
- Night Sight: Seeing in the Dark on Pixel Phones · Google Research · Primary source · 2018-11-14 · Exposure and frame count (retrieved 2026-11-08)
- Night Sight: Seeing in the Dark on Pixel Phones · Google Research · Primary source · 2018-11-14 · Exposure and frame count (retrieved 2026-11-08)
- [18]
Google said auto white balance fails in low light and developed a learning-based white-balance algorithm for Night Sight. confirmedas of 2018-11-14
- Night Sight: Seeing in the Dark on Pixel Phones · Google Research · Primary source · 2018-11-14 · Auto white balancing (retrieved 2026-11-08)
- Night Sight: Seeing in the Dark on Pixel Phones · Google Research · Primary source · 2018-11-14 · Auto white balancing (retrieved 2026-11-08)
- [19]
Typical digital zoom scales up a crop of a single image using interpolation, which tends to produce a blurry or plastic look lacking texture. confirmedas of 2018-10-15
- See Better and Further with Super Res Zoom on the Pixel 3 · Google Research · Primary source · 2018-10-15 · The Challenges of Digital Zoom (retrieved 2026-11-08)
- See Better and Further with Super Res Zoom on the Pixel 3 · Google Research · Primary source · 2018-10-15 · The Challenges of Digital Zoom (retrieved 2026-11-08)
- [20]
Google's Super Res Zoom on the Pixel 3 merged many handheld frames onto a higher-resolution grid, using natural hand tremor that shifts each frame by a few pixels, and Google said the detail was roughly competitive with 2x optical zoom lenses on many other phones. confirmedas of 2018-10-15
- See Better and Further with Super Res Zoom on the Pixel 3 · Google Research · Primary source · 2018-10-15 · Introduction (retrieved 2026-11-08)
- See Better and Further with Super Res Zoom on the Pixel 3 · Google Research · Primary source · 2018-10-15 · Practical Super-resolution Using Hand Motion (retrieved 2026-11-08)
- [21]
Apple's iPhone 17 Pro, announced in September 2025, has three 48MP rear cameras (Main, Ultra Wide and Telephoto), with 4x optical zoom at 100mm and an 8x optical-quality option at 200mm. confirmedas of 2025-09-09
- Apple unveils iPhone 17 Pro and iPhone 17 Pro Max · Apple Newsroom · Primary source · 2025-09-09 · Camera section (retrieved 2026-11-07)
- Apple unveils iPhone 17 Pro and iPhone 17 Pro Max · Apple Newsroom · Primary source · 2025-09-09 · Camera section (retrieved 2026-11-07)
- [22]
Apple says the iPhone 17 Pro's updated Photonic Engine uses machine learning in more parts of the image pipeline to preserve detail, reduce noise and improve color accuracy, especially in low light. confirmedas of 2025-09-09
- Apple unveils iPhone 17 Pro and iPhone 17 Pro Max · Apple Newsroom · Primary source · 2025-09-09 · Camera section (retrieved 2026-11-07)
Revision history (1)
- Page created.
Created Nov 8, 2026. Last reviewed by an editor on Nov 8, 2026. Next scheduled review: May 8, 2027.
Cite this page
"How phone cameras work: sensors and computational photography." ContentLora, updated Nov 8, 2026. https://contentlora.com/explain/how-digital-cameras-work
Spotted an error? Suggest a correction or emailcorrections@contentlora.com.