What is LOFIC? How smartphone cameras preserve highlights, and how it differs from other HDR methods
GDESIGN
DesignPublished: 14 min read

What is LOFIC? How smartphone cameras preserve highlights, and how it differs from other HDR methods

Expose for the view outside a bright window and the room looks dark. Expose for the room instead, and the sky may turn white. More megapixels alone cannot solve this: the camera has to record a wide range of light levels at once. LOFIC widens that range by storing charge that would otherwise overflow in bright light. Here is how it works, how it relates to other HDR methods, and what today's product claims actually tell us.

Japanese original published: 2026-11-17

The key idea: LOFIC does not repaint lost highlights

LOFIC stands for Lateral Overflow Integration Capacitor. The name describes a way to retain charge that a photodiode cannot hold: direct the overflow into additional storage capacitance, then use it in the image signal.

This operates at a different stage from darkening a white sky in an editing app. If differences in brightness were lost during capture, making the saved white pixels darker cannot recover the original cloud shapes from that data alone. LOFIC aims to reduce information loss during image capture.

It does not make highlight clipping impossible. The additional capacitance has a limit, as do readout and image processing. LOFIC is neither a guarantee of a particular phone's image quality nor a standard that makes camera-app features identical across devices.

Japanese source information checked November 16, 2026; English localization and cited sources checked November 17, 2026. The thumbnail is an AI-generated illustration of a high-contrast scene, not a camera sample or a performance comparison.

Why highlights clip: a pixel has finite charge capacity

A sensor pixel contains a photodiode that converts light into an electrical signal. Electrons generated during the exposure accumulate and are read out to form the image. For an intuitive picture, think of the electrons as dots and the storage capacity as the size of a container.

In dim light, the signal can be difficult to distinguish from noise, including noise introduced during readout. In strong light, storage or readout reaches its limit and further differences in brightness cannot be recorded. A bright cloud and an even brighter cloud can end up at the same maximum value, erasing the visible texture.

Figure 1. Keep overflow charge available for the signal

Reaching the storage limit

Store 5 of the 9 illustrative charges. If the remaining 4 cannot contribute to the signal, differences in bright light are lost.

Storing the overflow with LOFIC

Keep the 5 charges in the photodiode and store the 4 that overflow. Read out the stored charge to use it in the signal.

Containers and dots illustrate charge storage, not separate readout signals. They are not a physical cross-section, wiring diagram, capacity ratio or real electron count, and do not reproduce a particular product's circuit. Sources: 東北大学 工学研究科・工学部: 技術社会システム専攻の須川成利教授の研究グループが監視カメラ向けワイドダイナミックレンジCMOSセンサの実用化に成功 ~感度とダイナミックレンジでCCDを超える高画質を実現~; OMNIVISION: TheiaCel® Technology. Diagram or table by the NanToo editorial team, based on these sources.

LOFIC directs overflow charge into additional capacitance. The 5 and 4 in the diagram are invented teaching numbers, not real electron counts. The important point is not that there appear to be two containers: it is that charge that might otherwise be unavailable to the signal can be retained and read out.

Photodiode saturation is not the only cause of clipped highlights. Analog readout, conversion to digital values and final tonal adjustments can also reach their limits. EMVA 1288 distinguishes measured saturation capacity from the pixel's physical full-well capacity. Additional charge storage does not remove every downstream constraint.

HDR is not one method: multiple exposures, DCG and LOFIC

HDR means High Dynamic Range. It describes handling a wide range of brightness, not one exclusive capture method. The claim that older HDR always required a burst and LOFIC was the first way to use one exposure would be misleading.

A familiar approach combines a short exposure that preserves highlights with a longer exposure that gathers shadow information. This can work well for a static scene, but movement between exposures makes alignment harder. Canon's EOS R6 HDR mode, for example, merges three exposures and warns about afterimages from moving subjects. This is one implementation, not a rule that every smartphone combines exactly three frames.

Figure 2. Exposure timing versus reading at different gains

Multi-exposure HDR example

Short exposure: record highlights

Longer exposure at another time: record shadows

Align and merge images from different times

Single-exposure HDR with LOFIC

Accumulate charge during one exposure

Read charge from that exposure at different conversion gains

Combine the readout signals into an HDR image

Time runs horizontally in this conceptual comparison. Lengths and gaps are not product specifications. Multi-exposure implementations can also stagger timing by row; exposure and readout order depend on the implementation. Sources: Sensors: An Area-Efficient up/down Double-Sampling Circuit for a LOFIC CMOS Image Sensor; Canon: EOS R6: HDR Mode. Diagram or table by the NanToo editorial team, based on these sources.

Another approach reads signals from the same exposure at different conversion gains. DCG means Dual Conversion Gain. High conversion gain helps distinguish small amounts of charge; low conversion gain accommodates larger signals within the available readout range. This is different from digitally brightening a finished image.

Where charge is stored and which charge a readout signal includes are separate questions. In the representative circuit in the cited 2023 paper, the high-gain signal reads charge from the photodiode. The low-gain signal includes charge from the photodiode (PD), floating diffusion (FD, where charge is converted to voltage) and additional storage capacitance (CS). It does not read only the extra charge. Figure 1 does not mean making one image from 5 charges and another from 4, then simply adding the two images. Signals are combined with their different gains taken into account. Some designs share parts of the readout circuitry; not every product uses the same circuit.

Scroll horizontally if the table does not fit.

HDR methods can be combined, not just substituted for one another
MethodWhat changes or is addedLimits to consider
Multi-exposure HDRMerge images captured with different exposuresSubject movement, alignment and changes in illumination between exposures
Single-exposure HDR with DCGRead the same exposure at two conversion gainsPhotodiode charge capacity and the range of each readout
Single-exposure HDR with LOFICStore overflow charge and provide readout for itAdditional capacitance, circuit area, noise and signal-combination design
Hybrid methodsUse different gains together with multiple exposuresThe combination and supported shooting modes depend on the product
A comparison of principles, not an image-quality ranking. Sources: Sensors: An Area-Efficient up/down Double-Sampling Circuit for a LOFIC CMOS Image Sensor; Sony Semiconductor Solutions: Hybrid Frame-HDR(HF-HDR). Diagram or table by the NanToo editorial team, based on these sources.

LOFIC and DCG are not mutually exclusive: extra storage and readout at different gains can work together. Sony's HF-HDR also combines DCG-HDR with a very short exposure. A phone containing a single-exposure HDR technology can therefore still use multi-image processing for some photographs.

What does 100 dB mean? Not megapixels or file bit depth

Dynamic range describes the span between the dark-end detection threshold and bright-end saturation. The signal-to-noise ratio accepted at the dark end affects the result, so a number needs measurement conditions to be meaningful.

Figure 3. Dynamic range lies between two limits

Light increases from top to bottom

Below the detection threshold
Signal is difficult to distinguish from noise

Lower limit: absolute sensitivity threshold
SNR=1 in the EMVA definition

Dynamic range
Exposure at upper limit / exposure at lower limit

Upper limit: exposure at saturation

Beyond saturation
Further increases in light cannot be told apart

A conceptual view of the EMVA definition, not a quantitative scale. It does not guarantee usable image quality throughout the range. LOFIC primarily extends charge handling at the bright end, working with low-noise readout to provide a wide range. Sources: Sensors: An Area-Efficient up/down Double-Sampling Circuit for a LOFIC CMOS Image Sensor; EMVA: Standard for Characterization of Image Sensors and Cameras, Release 4.0 Linear. Diagram or table by the NanToo editorial team, based on these sources.

EMVA 1288 Release 4.0 Linear defines DR as the ratio of exposure at saturation to the absolute sensitivity threshold. Its decibel expression is 20 × log10(DR). The absolute sensitivity threshold is the mean number of incident photons at which the signal-to-noise ratio is 1 (SNR=1). At that point the signal equals the standard deviation of the noise. This is not a threshold for a sufficiently clean photograph.

Using that expression, 100 dB corresponds to 10(100/20) = 100,000, or log2(100,000) ≈ 16.6 stops. This is a unit conversion, not a measurement of a particular smartphone. It does not guarantee usable image quality across all 16.6 stops.

Scroll horizontally if the table does not fit.

Converting dB into ratios and stops
Dynamic rangeRatioStops, approximately
60 dB1,000:110.0
80 dB10,000:113.3
100 dB100,000:116.6
Arithmetic examples using 20 log10(DR). One stop represents a factor of two. These are not comparisons between sensors. Sources: EMVA: Standard for Characterization of Image Sensors and Cameras, Release 4.0 Linear. Diagram or table by the NanToo editorial team, based on these sources.

Sony's 100 dB figure for L910 is a manufacturer specification. It can be converted with the formula above, but we have not established that this product is EMVA-compliant. Comparing dB figures from different manufacturers also requires checking exposure count, readout mode and noise-evaluation conditions.

A 50-megapixel count describes how many pixels sample the image. A 12-bit value describes a digital code's bit depth. Dynamic range describes a span of signals. They are related, but not interchangeable: 16.6 stops does not mean a 16.6-bit file, and a higher pixel count alone does not prove better highlight retention.

Not a new invention: from the 2000s to smartphones

LOFIC has recently become more prominent in smartphone discussions, but its foundations are older. Tohoku University's 2008 announcement describes joint development with Texas Instruments Japan starting in 2003, followed by research presentations from 2005. That release also reported commercialization of TC922 for surveillance cameras.

Scroll horizontally if the table does not fit.

Milestones supported by the cited primary sources
PeriodDocumented developmentHow to interpret it
From 2003Tohoku University and Texas Instruments Japan collaborate; research presentations follow from 2005The underlying idea did not suddenly appear in 2026
September 2008Commercialization of TC922 for surveillance cameras is announcedThis was a different application from today's smartphone sensors
March 2024OMNIVISION announces OV50K40 with TheiaCelA smartphone-oriented product; the current product page lists mass production
June 2026Sony announces LYTIA L910 with LOFICFirst in the LYTIA lineup, not the first LOFIC sensor worldwide
Selected documented milestones, not a complete history of the invention. Sources: 東北大学 工学研究科・工学部: 技術社会システム専攻の須川成利教授の研究グループが監視カメラ向けワイドダイナミックレンジCMOSセンサの実用化に成功 ~感度とダイナミックレンジでCCDを超える高画質を実現~; OMNIVISION: OMNIVISION Launches Industry’s First and Only Image Sensor with TheiaCel® Technology for Best-in-Class HDR in High-End Smartphones; OMNIVISION: OV50K40; Sony Semiconductor Solutions: Sony Semiconductor Solutions to Release LOFIC-Equipped CMOS Image Sensor for Mobile Applications with 100 dB Dynamic Range and Low Power Consumption. Diagram or table by the NanToo editorial team, based on these sources.

Why does development take so long? Additional storage and low-noise readout consume circuit area and engineering effort. With many small pixels packed together, simply adding a large capacitor is not the end of the design problem. The 2023 readout-circuit paper explicitly addresses area efficiency.

The useful interpretation is not that HDR was impossible until now. It is that designs have progressed toward combining a wide capture range with compact pixels, high resolution and video. Separating research, sensor commercialization and adoption in finished devices makes the news easier to assess.

Product examples: LYTIA L910 and phones that specify LOFIC

Sony describes LYTIA L910 as combining LOFIC with TCG-HDR, or Triple Conversion Gain-HDR: charge from one exposure is read at three conversion gains. The relevant design is the combination of storage, readout and noise control, not the LOFIC name in isolation.

The announced specifications include approximately 50 effective megapixels, a 1/1.28-type format, 1.22-micrometer pixel pitch, 100 dB from a single exposure and HDR video at 4K/60 fps. Full-resolution readout and 4K video are different modes; the 4K/60 fps specification uses 2 × 2 binning. Do not combine every maximum into one operating mode.

The announcement scheduled mass-production shipments for summer 2026. A schedule alone does not establish actual shipments today or identify the sensor in a particular phone.

For a finished-device example, Xiaomi's global Xiaomi 17 Ultra page explicitly lists LOFIC HDR for its Light Fusion 1050L main camera. This confirms a claimed feature at the phone level, not just a sensor announcement. It does not prove that the phone uses L910 or that its ultra-wide and telephoto cameras use the same method.

These are examples, not a recommendation ranking or a measure of market share. Check regional specifications, warranty coverage, the camera you intend to use and its video modes. We have not conducted hands-on comparisons and do not label a product the best based on manufacturer marketing.

The limits: one exposure does not eliminate blur or distortion

Single-exposure HDR can avoid the alignment problems caused by merging captures from different times. But one exposure still has a duration. If the subject or camera moves during it, the image can blur. Misalignment between exposures and motion blur within an exposure are different problems.

Single exposure also does not mean global shutter. A rolling shutter captures different rows at different times, which can distort moving objects. OV50K40, despite using LOFIC, explicitly lists Rolling Shutter in its specifications. One exposure for HDR does not imply simultaneous capture across all pixels.

Additional capacitance does not create missing light. A bright sign beside a dark sidewalk is a different problem from a scene that is dark everywhere. In the latter case, collected light, exposure duration and low-noise readout remain important. Better highlight handling should not be expanded into a promise that every aspect of low-light quality improves.

Flicker from LEDs and similar lighting also depends on the light's timing and the exposure/readout sequence. A manufacturer's reduction claim under particular conditions is not a guarantee that banding disappears under every light source.

Figure 4. LOFIC is one part of the finished photograph

1. Optics and exposure
How much light is collected, and when.

2. Pixels and readout
Store charge and obtain signals. This is where LOFIC operates.

3. Image processing
Combine signals, reduce noise, and adjust color and tones.

4. Storage and display
Deliver an image file and view it on a screen.

A simplified photography workflow. The curve represents the idea of tonal adjustment, not a measured product response. Sources: Sensors: An Area-Efficient up/down Double-Sampling Circuit for a LOFIC CMOS Image Sensor; Canon: EOS R6: HDR Mode; Sony Semiconductor Solutions: Hybrid Frame-HDR(HF-HDR). Diagram or table by the NanToo editorial team, based on these sources.

There are further choices in how the captured range is presented. Preserving highlight gradations while lifting shadows by different amounts changes a photograph's appearance. Sensor design is a foundation, but optics, capture control, processing and the final image's presentation all deserve separate consideration.

What to check when choosing a camera, and when shooting

LOFIC is worth understanding if you often photograph interiors with bright windows, backlit people or night scenes combining bright lights and dark backgrounds. Whether a specific phone improves those scenes still requires comparable photographs and shooting-mode information. The following is an editorial checklist derived from the principles, not the result of our own device tests.

  • Which camera: Is the feature limited to the main camera? Do the telephoto and ultra-wide cameras you use have different specifications?
  • Which mode: Is it available for both stills and video, and at the resolution and frame rate you need?
  • Moving subjects: Look beyond static night scenes. Are walking people and moving leaves rendered naturally?
  • Shadows and highlights together: If lettering on a bright sign remains visible, do shadows become noisier? Does a brighter result conceal lost fine detail?
  • Comparable conditions: Match field of view, lighting, display size and output format. Do not rank cameras by dB alone.

On your current camera, lowering exposure may preserve highlights at the expense of shadows. A shorter exposure can help with movement, but gathers less light. LOFIC does not remove these choices; it can expand the room available for making them.

The point of LOFIC is to lose less information at capture, not to invent missing differences in brightness afterward. It improves charge storage and readout rather than magically replacing every HDR technique. Read the method and its operating conditions together to see what megapixel counts alone cannot tell you.

For the next stage, see our guide to JPEG quality and compression. Information acquired during capture and information preserved during file storage are separate questions.

References and sources

Editorial note

This article was prepared with AI assistance and reviewed by an editor before publication. It may still contain factual errors, interpretation mistakes, or outdated information. Check the cited primary sources or official documentation before making an important decision.

Related articles