{
  "id": "rolling-shutter",
  "title": "Rolling Shutter",
  "lang": "en",
  "source_lang": "en",
  "summary": "Rolling shutter is an image-capture method that exposes different parts of a frame at different times, potentially producing geometric distortion or brightness bands when motion or lighting changes during capture.",
  "category": "technology",
  "subcategory": "engineering",
  "type": "concept",
  "aliases": [
    "Electronic rolling shutter",
    "Rolling-shutter capture"
  ],
  "generated_at": "2026-11-05",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/rolling-shutter/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/rolling-shutter/",
      "json": "https://aiwiki.page/zh/rolling-shutter.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "A rolling shutter is an image-capture method in which different parts of an image are exposed at different times, usually through a sequence of sensor rows. A recorded frame therefore represents a short interval of time rather than one shared exposure interval across the entire image. Rolling shutters are common in cameras using CMOS image sensors, including smartphones, video cameras, and industrial cameras. They can produce characteristic distortions when the camera or subject moves, or brightness bands when illumination changes during the scan."
    },
    {
      "heading": "Operating principle",
      "text": "An image sensor converts incident light into electrical signals. In a typical electronic rolling shutter, each row of light-sensitive pixels is reset to begin exposure, accumulates charge for a specified interval, and is then sampled. These operations proceed across the sensor with a delay between successive rows.\n\nIndividual rows usually have overlapping exposure intervals; the process does not normally mean that one row finishes its entire exposure before the next begins. Nevertheless, the start and end times differ from row to row.\n\nFor a simplified top-to-bottom scan, the exposure start time of row \\(y\\) can be written as\n\n\\[t_{\\mathrm{start}}(y)=t_0+y\\tau,\\]\n\nwhere \\(t_0\\) is the first row’s start time and \\(\\tau\\) is the delay between successive rows. If every row has an exposure duration \\(T_e\\), then\n\n\\[t_{\\mathrm{end}}(y)=t_0+y\\tau+T_e.\\]\n\nFor an image containing \\(H\\) rows, the first-to-last-row timing difference is approximately\n\n\\[T_s=(H-1)\\tau.\\]\n\nThis interval, often called the scan time or rolling-shutter readout time, is central to the severity of rolling-shutter artifacts. Actual sensors may use different scan directions, multiple readout regions, or more complicated timing patterns.\n\nA simple model of the recorded image is\n\n\\[I(x,y)\\propto \\int_{t_0+y\\tau}^{t_0+y\\tau+T_e} E(x,y,t)\\,dt,\\]\n\nwhere \\(E(x,y,t)\\) denotes the irradiance reaching sensor position \\((x,y)\\) at time \\(t\\). The row-dependent limits distinguish this model from simultaneous exposure.\n\nExposure time, scan time, and frame rate\n\nThree quantities must be distinguished:\n\n- Exposure time: how long each row collects light.\n- Scan time: the timing difference between the first and last rows.\n- Frame interval: the time between successive frames.\n\nA short exposure can reduce motion blur without eliminating rolling-shutter distortion. For example, each row might be exposed for only a small fraction of a millisecond while the complete scan takes several milliseconds. Moving objects can then appear sharply defined but geometrically deformed.\n\nLikewise, frame rate alone does not uniquely specify scan time. Cameras operating at the same frame rate can have different rolling-shutter performance."
    },
    {
      "heading": "Characteristic artifacts",
      "text": "Skew and stretching\n\nWhen an object moves horizontally during a vertical scan, different rows record it at different horizontal positions. A vertical pole may therefore appear tilted, and a rectangular object may become a parallelogram.\n\nUnder an idealized constant horizontal image velocity \\(v_x\\), measured in pixels per second, the first-to-last-row displacement is approximately\n\n\\[\\Delta x=v_xT_s.\\]\n\nThis relation describes the timing-induced shear, not all possible effects of camera rotation, perspective, or changing depth.\n\nMotion along the scan direction can instead make objects appear stretched or compressed. The result depends on the direction and speed of motion relative to the scan.\n\nWobble and vibration\n\nRapid changes in camera orientation during a scan can bend straight lines or produce an undulating appearance commonly called the jello effect. This is especially noticeable in footage from vibrating vehicles, drones, and lightly mounted cameras.\n\nA constant pan tends to produce a more regular skew, whereas vibration can generate a distortion that changes repeatedly within a single frame.\n\nRotating-object distortion\n\nPropellers, fans, wheels, and other rotating objects can appear curved, broken, or disconnected. Each sensor row records the rotating object at a different angular position, so the assembled image need not resemble any instantaneous configuration of the object.\n\nThese effects can coexist with temporal aliasing, such as the apparent reversal of a rotating wheel. The two phenomena are distinct: rolling shutter introduces timing differences within a frame, while frame-to-frame sampling can create ambiguous apparent motion even with simultaneous exposure.\n\nBrightness bands and partial flash exposure\n\nIf illumination varies during the scan, rows receive different amounts of light. Flickering lamps, pulse-driven displays, and pulse-width-modulated lighting can therefore create bright and dark bands.\n\nA brief photographic flash may illuminate only those rows whose exposure intervals overlap the flash. Part of the frame can consequently appear brightly lit while the rest remains dark. Band orientation and movement depend on scan direction, exposure duration, illumination frequency, and timing."
    },
    {
      "heading": "Comparison with other shutter systems",
      "text": "Global shutter\n\nA global shutter gives all pixels approximately the same exposure interval. Their stored signals may still be read sequentially afterward. Sequential data output alone therefore does not make a sensor a rolling-shutter sensor: the decisive issue is when the light contributing to each pixel is captured.\n\nGlobal shutters substantially reduce the geometric artifacts caused by row-dependent exposure timing. They do not eliminate ordinary motion blur, lens distortion, or frame-to-frame temporal aliasing.\n\nGlobal-shutter implementations often require additional pixel circuitry or storage. Depending on sensor design and manufacturing technology, this can introduce trade-offs in pixel size, cost, sensitivity, noise, or dynamic range. These are design-dependent comparisons rather than universal disadvantages.\n\nMechanical shutters\n\nRolling exposure is not exclusive to electronic sensors. A mechanical focal-plane shutter can expose a photographic image through a moving slit, particularly at short exposure settings. Different parts of the image then receive light at different times, allowing similar motion distortions.\n\nAn electronic first-curtain shutter combines an electronic exposure start with a mechanical exposure end. Its timing characteristics depend on the particular implementation. The label “mechanical shutter” therefore does not necessarily imply simultaneous exposure across the frame."
    },
    {
      "heading": "Development and use",
      "text": "Moving-slit exposure predates digital photography. Electronic rolling shutters became especially widespread as CMOS sensors gained importance in digital cameras, webcams, smartphones, and video systems. Their prevalence reflects their compatibility with practical CMOS pixel and readout architectures, rather than an inherent requirement that all CMOS sensors use rolling exposure.\n\nRolling-shutter sensors are used in consumer photography, cinematography, microscopy, machine vision, and robotics. For static scenes under stable illumination, their timing differences may have little visible effect. For fast motion or precision geometric measurements, those differences can become important.\n\nIn remote sensing and aerial imaging, camera movement during exposure can alter the geometry needed to infer positions and dimensions. In robot navigation, treating every pixel in a frame as if it were captured at one instant can bias estimates of camera motion and scene structure."
    },
    {
      "heading": "Mitigation and correction",
      "text": "Capture-side measures\n\nA shorter scan time generally reduces rolling-shutter distortion. Faster sensor readout, more parallel readout circuitry, and some reduced-resolution or cropped modes can shorten the scan, although the effect depends on camera design.\n\nReducing camera vibration or relative subject motion can also reduce artifacts. Shortening exposure chiefly reduces blur; it does not necessarily shorten the row-to-row timing difference. For changing illumination, exposure settings and lighting synchronization may reduce banding, but performance depends on the light waveform and camera timing.\n\nComputational correction\n\nCorrection algorithms estimate how the camera or scene moved during the scan and then remap image coordinates. They may use image features, optical flow, or measurements from a gyroscope. Combining image and inertial measurements is an application of sensor fusion.\n\nReliable correction may require calibration of scan direction, row timing, camera geometry, and synchronization between the camera and motion sensors. A correction based only on camera rotation cannot fully account for independently moving objects or depth-dependent displacement caused by camera translation.\n\nPost-processing also has limits. Occluded surfaces, clipped highlights, missing flash illumination, and detail lost to blur cannot generally be recovered by geometric warping alone. Conversely, rolling-shutter-aware vision algorithms can incorporate row-specific capture times directly into their estimation models rather than first attempting to produce a globally exposed image."
    },
    {
      "heading": "Measurement and specification",
      "text": "Useful rolling-shutter specifications include scan time, scan direction, operating resolution, cropping mode, and frame rate. A single sensor may have substantially different timing in different recording modes.\n\nScan time can be measured using controlled motion or illumination with known timing. Comparisons must specify whether “readout time” means the first-to-last-row exposure offset, active sensor readout, or a larger interval that includes blanking and data transfer. These quantities are related but are not always interchangeable."
    }
  ],
  "links": [
    "cmos-image-sensor",
    "aliasing",
    "modulation",
    "global-shutter",
    "robotics",
    "remote-sensing",
    "optical-flow",
    "gyroscope",
    "sensor-fusion",
    "calibration"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}