Camera

A camera is a device that forms an image by focusing light from a scene onto a light-sensitive surface and recording the result. Its basic design, a light-tight enclosure with an opening that admits light through a lens, has changed little in principle since the early 19th century, while the recording medium has moved from chemically treated plates and film to electronic sensors. Modern cameras range from camera obscura-derived optics in large-format studio instruments to the multi-lens systems in smartphones, and they are used for photography, cinema, science, medicine, industry, and machine perception. Because cameras turn light into data, they connect optics, chemistry, electronics, and computing.

The article covers the physics of image formation and lens design, the historical progression from the camera obscura through photochemical processes to digital imaging, and the engineering of sensors, exposure control, focusing, and image processing. It then surveys the main camera types, motion-picture and video recording, camera phones and computational photography, and specialised scientific and industrial instruments. The final sections address image quality and its physical limits, the use of cameras in computer vision and artificial intelligence, the economic transformation of the imaging industry, and the social, legal, and ethical questions raised by pervasive image capture.

Basic Principles of Image Formation

A camera forms an image by admitting light from a scene through a small opening and projecting it onto a surface. In its simplest form, the camera is a box with a single hole. Light from the scene travels in straight lines through the opening and strikes the opposite wall, where it forms a real image that can be seen on a translucent screen or recorded by a sensor. The enclosure must be light-tight. Any stray light that enters by another route falls across the image as a fairly uniform haze, which lowers contrast and makes the picture look washed out.

Each point of the scene sends light in many directions, but the opening passes only the narrow bundle of rays that travel along the line joining that point to the hole. Each scene point is therefore mapped to one corresponding point on the back surface. Rays from the top of the scene reach the lower part of the image, and rays from the bottom reach the upper part, so the image is inverted both vertically and horizontally. The inversion follows from the straight-line geometry of the rays and does not require any focusing element. Moving the image surface closer to the opening shrinks the image in proportion to the distance, because the ray geometry scales with the distance between the hole and the surface.

The opening is never a mathematical point, and its size determines the trade-off between sharpness and brightness. A hole of diameter $d$ turns each scene point into a small spot of roughly $d(u+v)/u$ on the image plane, where $u$ is the distance to the object and $v$ is the distance from the hole to the image plane. A smaller hole gives sharper detail but admits less light, and a larger hole gives a brighter but blurrier image. Below a certain size, however, the diffraction of light, a wave effect, spreads each point out again, so shrinking the hole indefinitely does not keep improving sharpness.

Image brightness is governed by irradiance, the light power received per unit area of the image surface, and the exposure is the irradiance multiplied by the time during which the surface receives light. For a pinhole, the admitted light is proportional to the hole's area, while the image area grows with the square of the image distance. The irradiance is therefore roughly proportional to $(d/v)^2$, or $1/N^2$, where $N = v/d$ is the pinhole's f-number. This is why a pinhole image is dim, and why halving the hole diameter reduces the brightness to a quarter. Because the rays from each point are limited to a narrow cone, the pinhole also keeps blur spots small over a wide range of object distances, so most of the scene appears sharp at once.

A lens solves the brightness problem by collecting light from a wide cone and bending it so that the rays from each object point reconverge at one image point. This relation between object distance, image distance, and focal length is described by the thin-lens equation, $1/f = 1/u + 1/v$. For a fixed image distance, only objects at one matching distance are rendered as sharp points; objects nearer or farther produce blurred disks instead. This is the reason a lens-based image is sharp only for objects at certain distances, and the pinhole's tolerance for varying distances is the price paid for its dimness. Both designs rest on the same underlying rules of light propagation and geometry described in optics.

Lenses and Optical Design

Focal length is the basic parameter of a lens. For a distant subject, it is the distance from the lens to the plane where the image forms, and it fixes both image scale and angle of view. For a rectangular sensor, the horizontal angle of view is $2\arctan\left(\frac{w}{2f}\right)$, where $w$ is the sensor width and $f$ the focal length. On a full-frame sensor about 36 mm wide, a 50 mm lens covers roughly 40 degrees horizontally, while a 200 mm lens covers about 10 degrees. Longer focal lengths enlarge distant subjects but narrow the field.

For a lens whose thickness is small compared with its focal length, the optical relationship between object distance $d_o$, image distance $d_i$ and focal length $f$ follows the thin-lens approximation: $\frac{1}{f} = \frac{1}{d_o} + \frac{1}{d_i}$. Lateral magnification is $m = -\frac{d_i}{d_o}$, and the negative sign shows that the image is inverted. A 50 mm lens focused on a subject 2 m away forms its image about 51.3 mm behind the lens, at a magnification of roughly −0.026. Real lenses are compound systems whose effective focal length is measured from principal planes, but the approximation underlies most first-order design. It follows from applying Snell's law to each curved surface, which is the refraction that bends rays toward a focus. Focusing on closer subjects requires the image distance to increase, which is why the lens moves away from the sensor.

The f-number, $N = \frac{f}{D}$, is the ratio of focal length to the aperture diameter $D$. Illuminance at the image plane scales with $1/N^2$, so each full stop multiplies $N$ by about $\sqrt{2}$ and halves the light. Depth of field grows with $N$: for subject distances $u$ much greater than $f$, the total depth is approximately $\frac{2Nc\,u^2}{f^2}$, where $c$ is the acceptable circle of confusion. Small apertures therefore increase depth of field at the cost of light, and at very small apertures diffraction limits sharpness.

Real lenses deviate from ideal imaging through optical aberrations. Spherical aberration occurs when rays passing through the edge of a lens focus at a different point from central rays; it is reduced by stopping down, by choosing suitable surface shapes, and by aspherical elements. Chromatic aberration arises because the refractive index varies with wavelength, a property known as dispersion, so different colours focus at different distances and magnifications. Achromatic doublets pair crown and flint glasses with different dispersion to cancel much of this error, and apochromatic designs add low-dispersion elements. Coma makes off-axis point sources appear comet-shaped, and distortion bends straight lines into barrel or pincushion shapes because magnification changes with distance from the centre. Field curvature and astigmatism also degrade the edges of the frame. Designers use ray-tracing software to optimise multi-element groups in which errors of opposite sign cancel, balancing sharpness against size, weight and cost.

Surface coatings address losses and stray light. An uncoated glass surface with a refractive index near 1.5 reflects about 4% of normal incident light, and these losses accumulate across many surfaces while reflections cause flare and ghost images. Anti-reflection coatings are thin layers, typically near a quarter wavelength thick, whose reflected waves interfere destructively; multilayer coatings extend this effect across the visible spectrum. This is a practical application of reflection and interference.

Compact designs rely on specialised components. Aspherical elements, whose surfaces are not spheres, correct spherical aberration with fewer glass elements, reducing size and weight; moulded glass and plastic aspherics are common in mass-produced and smartphone lenses. High-refractive-index glass allows stronger curvature and therefore shorter lenses with less aberration, while low-dispersion glass reduces chromatic error. Together these methods allow small lenses to reach sharpness that once required much larger optical assemblies.

Camera Obscura and Early Optical Knowledge

The earliest surviving accounts of the camera obscura come from China and Greece. The earliest documented explanation of this principle comes from Chinese philosopher Mozi, who correctly argued that the inversion of the image results from light travelling in straight lines. The dating of Mozi's writings is disputed: one source places them around 500 BCE, while another dates the text to the 4th century BC. Later, in 330 B.C.E., the Greek philosopher Aristotle noticed a partial solar eclipse projected on the ground through the holes in a sieve and the gaps in tree foliage. Aristotle's observation is often cited as an early instance of the same effect, in which many small openings each produce an image of the sun.

The most detailed early study came from Ibn al-Haytham (Alhazen), whose Book of Optics was written in the early 11th century. The Book of Optics is a seven-volume treatise on optics that argued against the extramission theory of vision and proposed the modern intromission theory, the now accepted model that vision takes place by light entering the eye. He worked in a darkened room he called Al-Beit Al-Muzlim, which was later rendered in Latin as camera obscura. In that room he observed the light of the Moon passing through a small hole and asked why the projection appeared upside down. That question led him to the conclusion that light moves in straight lines. He also noticed that smaller holes would create sharper images. His work was also methodical: his method consisted of a cycle of observation, hypothesis, experimentation, and independent verification. Some popular accounts credit him as the inventor of the device, but historians are more cautious; the history of photography notes he is said to be the first that actually built a camera obscura.

In the Renaissance, the device moved from the study of optics into the studio. Leonardo da Vinci's notebooks, including the Codex Atlanticus and Manuscript D, contain the earliest known diagrams and detailed descriptions of the pinhole camera, dating from about 1490. Leonardo recognised that because the light rays crossed as they passed through the hole, the projected image was inverted or upside down. Improvements followed in the 16th century. A biconvex lens in the opening (first described by Gerolamo Cardano in 1550) and a diaphragm restricting the aperture (Daniel Barbaro in 1568) gave a brighter and sharper image. Giambattista della Porta, in his widely read books, was the first to write a description of using the camera obscura as a drawing aid in 1558. Artists used the projected image to check proportions and the image created by the camera obscura is an image in true perspective, meaning that the image maps the world's three dimensions onto a two-dimensional surface in a mathematically precise fashion. This made the device a practical support for linear perspective. Painters such as Johannes Vermeer and Canaletto utilized the camera to enhance their perspective and accuracy in painting. By the late 17th century, portable camera obscura devices in tents and boxes had come into use as drawing tools.

The device had a fundamental limitation: it could show an image but not keep it. The images produced by these early cameras could only be preserved by manually tracing them, as no photographic processes had been invented yet. The trade-off between sharpness and brightness was also unresolved. The smaller the hole, the sharper but dimmer the image. Lenses improved brightness but introduced their own focusing problems, which is why later users adjusted the lens position to bring the image into focus. The sharpest image could be seen on a screen, yet nothing on that screen could be stored. This gap made the fixing of images the central problem for the inventors of photography. Later designers adapted the boxes directly: such cameras were later adapted by Louis Daguerre and William Fox Talbot for creating the first photographs.

The name itself is also a point of disagreement. One source states that the term was coined by the astronomer Johannes Kepler in the seventeenth century, while other popular accounts attribute the coinage to Ibn al-Haytham. The historical record supports the earlier date for the optical principle rather than a single originator of the term.

Fixing the Image: Early Photochemical Processes

A camera obscura could project a scene, but nothing held the projection in place, so the central problem of the 1820s to 1850s was to make a projected image chemically durable. Nicéphore Niépce's early work on silver chloride paper, around 1816, captured small camera images, but he could not stop them darkening when they were brought into the light to be viewed. He made the first apparent successes with silver chloride on paper, but he could not fix the image permanently; it gradually darkened overall. He then turned to bitumen of Judea, a natural asphalt that engravers already used as an acid-resistant coating, and he worked with it in the 1820s. The material, initially soluble in spirits and oils, was hardened and made insoluble in the brightest areas of the image. Rinsing away the unhardened part left a durable record, a process he called heliography.

Niépce applied this method to the camera. His 1826 or 1827 view from his estate at Le Gras, on a pewter plate, is the oldest surviving camera photograph. The traditional estimate of about eight hours of exposure comes from the sunlight on the buildings, but a researcher who recreated his process concluded that the exposure must have lasted several days. Such exposures were impractical, and after Niépce's death in 1833 Daguerre concentrated his attention on the light-sensitive properties of silver salts.

Daguerre's process, announced in 1839, changed the approach to chemistry. He exposed a thin silver-plated copper sheet to iodine vapour, which formed a coating of light-sensitive silver iodide. The plate was exposed in a camera, and the key discovery was that an invisibly faint "latent" image created by a much shorter exposure could be chemically "developed" into a visible image. The latent image was developed with the vapour given off by mercury heated to 75 °C. Sources differ on the exposure time: one describes plates exposed for over 20 minutes, while another gives exposure times ranging from a few seconds to several minutes, depending on the lighting conditions. The French government offered Daguerre a lifetime pension in return for the details, and he disclosed the method in August 1839. Each daguerreotype was a single, non-reproducible image, and it was far superior to the calotype in detail and far more popular in its early years.

William Henry Fox Talbot pursued a different route based on negatives. He published on paper negatives in 1835, and these so-called photogenic drawings were basically contact prints on light-sensitive paper, which unfortunately produced dark and spotty images. His improved process, the calotype, was modified and named the calotype in 1840, and unlike the original process, it used a much shorter exposure time and a development process following exposure. Talbot patented the process in 1841, and this patent is one of the early examples of a patent on a photographic method. Because the calotype gave a negative, Talbot could generate multiple positive prints of a single image. This reproducibility, rather than the sharper detail of the daguerreotype, became the foundation for later photography.

Later refinements aimed at shorter exposures and easier handling. Frederick Scott Archer's wet collodion process, in the early 1850s, obtained an emulsion on a glass plate from an iodised solution of pyroxylin dissolved in alcohol and ether, which formed a transparent coating when the solvents evaporated. The plate had to be sensitised and exposed while still wet, which limited its use in the field, but it gave much faster exposures than the earlier methods. Dry gelatin plates, which followed from the 1870s, removed the need for immediate processing and made photography practical for both professionals and amateurs.

Film Cameras and the Rise of Consumer Photography

Through the nineteenth century, most photographers worked with heavy, fragile glass plates that had to be prepared and loaded by hand. Eastman not only invented roll film in 1884—that is, a strip of photographic film wound into a roll that could easily be inserted into a camera, and this flexible roll film offered a lighter alternative to plates. On September 4, 1888, inventor George Eastman secures U.S. Patent No. 388,850 for a small roll-film hand camera and registers the trademark "Kodak", and that patent-protected design brought the new medium to ordinary buyers. Its appeal lay as much in the business model as in the hardware.

The camera came preloaded with a 100-exposure roll of flexible film. Buyers did not develop anything themselves. Once the preloaded 100-exposure roll was finished, the camera could be mailed to Eastman's factory in Rochester, New York. The company then developed the film, returning mounted prints, negatives and the camera reloaded with fresh film to the owner. Eastman condensed this into a promise that the customer only had to press the button while the company handled the rest. The camera was sold for $25 with the film sealed inside. Each exposure produced a circular picture roughly 2.5 inches (6 cm) in diameter. Demand was strong, with buyers taking up more than 5,000 units within a year. The popularity led the firm to adopt the Eastman Kodak name in 1892 (The immediate triumph of the camera prompted Eastman to change the name of his company from Eastman Dry Plate and Film Company to Eastman Kodak Company in 1892.). Because no chemical or darkroom skill was needed, the photographer shifted from a specialist craft to a consumer activity.

A second change came from cinema. Between 1913 and 1914 he developed the Ur-Leica, a prototype camera using 35mm motion picture film. The 24 × 36 mm still frame was obtained by doubling the normal movie-image. The First World War interrupted commercial plans, and it wasn't until 1925 that the first Leicas were made commercially available. Kodak later named the format "135", as coined by Kodak in 1934. The format's influence then extended well beyond Leitz: its adoption of 35mm film for still-camera use became standard for Leitz's commercially available Leica (Leitz Camera) models that followed — and later for the still-photography industry at large. This kept the same film size in use across generations of cameras.

The third development was instant photography. Edwin Land, an American inventor known for inexpensive polarizing filters, began this work after in 1943, Land made another breakthrough, inspired by his daughter. She asked him why she couldn't see photos right after they were taken. The public debut came on Feb. 21, 1947, at a meeting of the Optical Society of America. The Model 95 reached shops on 26 November 1948: the first "Land Camera"—better known today as the instant Polaroid camera—goes on sale at Jordan Marsh department store in Boston for $89.75. It sells out within minutes that first day.

The process packed darkroom steps into the film itself. When the film was released from the camera, a pair of rollers at the mouth of the camera bit the pod, rupturing it and allowing reagent to be forced evenly across the film, coating the entire image area. As the reagent spread, various chemicals worked to remove the unexposed silver halide from the negative, release it onto the positive layer at the top of the film unit, and reduce it, producing the final image. It produced a sepia photograph in about one minute. Colour followed much later, with the introduction of the world's first instant color film in 1963.

Each instant print was a unique original, and every exposure carried a real cost in film and processing. These limits shaped the economics of film-era photography, which remained the default until electronic sensors began to replace chemical media (see photography and chemistry for the underlying principles).

Electronic Image Sensors: CCD and CMOS

A digital image sensor is a grid of light-sensitive sites built on a semiconductor, usually silicon. A photon that is absorbed can free an electron through the photoelectric-effect, so each pixel accumulates electric-charge in proportion to the light it receives during an exposure. The engineering problem is reading out millions of these small charges quickly, accurately and cheaply. The two main architectures, the charge-coupled device (CCD) and the CMOS sensor, answer that problem in different ways.

The charge-coupled device

The CCD did not begin as an imaging device. Willard Boyle and George Smith, working at Bell Labs in 1969, were trying to build an electronic substitute for magnetic bubble memory. While working at Bell Labs in 1969, Boyle and Smith sketched out the basic CCD in about an hour, and built a working prototype in under a week. Their CCD work was then developed specifically for imaging applications by Michael F. Tompsett, also at Bell Labs, in the 1970s. The imaging use is what led to the 2009 Nobel Prize in Physics for Boyle and Smith.

In operation, the charge is held in discrete packets in columns of closely spaced semiconductor capacitors. Stored information is read by shifting stored charges down the columns, one position at a time. Charge from the last element is sent to a single output amplifier, which turns it into a voltage. Because all pixels share that one amplifier, the device tends to produce uniform images with low noise, but it needs more power and reads out more slowly than its competitor.

The CMOS sensor

The CMOS sensor takes the opposite approach. Each pixel in a CMOS sensor contains not only a photodiode but also a small amplifier, as well as other circuitry, that converts the accumulated charge into a voltage signal right within the pixel itself. Because the conversion happens in parallel across the array, the readout can be fast. Since only one row of pixels at a time needs to be active during the readout and there is no charge transfer (and associated switching) from pixel to pixel during image acquisition, power consumption is low, which matters for batteries.

CMOS sensors are also cheap to make. CMOS image sensors can be made at standard manufacturing facilities that produce almost 90% of all semiconductor chips, which significantly decreases their design and manufacturing costs. Because the photodetector and the processing circuits share one chip, camera-control logic and analogue-to-digital conversion can be integrated alongside the pixels. Early CMOS imagers were weaker in sensitivity, dynamic range and linearity, but comparable in modern sensors due to advancements in technology. By the late 1990's CMOS image sensors began replacing CCD's in less demanding imaging applications. Today you likely have a CMOS sensor in your smartphone and consumer-level camera.

Kodak's 1975 prototype

The first self-contained digital camera was built at Kodak in December 1975 by Steven Sasson. It used a CCD with a 100 × 100 pixel array, which the Fairchild chip at the centre of the device supplied. It weighed 8 pounds with its toaster-size body, and took 23 seconds to capture an image of 0.01 megapixels on a cassette tape. The sensor itself was fast: the CCD captured an image in about 50 milliseconds. The slow part came afterward. Data moved into temporary memory and then onto cassette tape, leaving the camera unavailable for another capture until the transfer finished.

The prototype showed that a camera could work without film, but it was not a product. Management was reluctant to pursue it, since the company's business rested on film. Commercial electronic still cameras appeared only later; the first electronic still cameras to be marketed were shown as early as 1981, when Sony demonstrated the prototype Pro Mavica.

Colour Capture and Image Reconstruction

Colour photography grew out of research on human vision. In 1855 James Clerk Maxwell proposed that a full-colour scene could be recorded as three monochrome images, each made through a different filter, and later combined to reproduce the colours. In 1861 he tested the idea at the Royal Institution. The photographer Thomas Sutton made three negatives of a tartan ribbon through red, green and blue filters, and the resulting positives were projected through matching filters until they overlapped into a single coloured picture. Maxwell's principle of recording three channels still underlies almost every colour process, from film to digital sensors.

Digital sensors do not measure colour directly. Each photosite responds to the total intensity of the light reaching it, so manufacturers place a colour filter array over the pixels. The most common arrangement is the Bayer pattern, a repeating 2×2 block containing two green filters on one diagonal and one red and one blue filter on the other. The doubled green reflects the eye's greater sensitivity to that part of the spectrum. Because of this layout, only one pixel value among RGB values is recorded at each pixel, and the image that leaves the sensor is a mosaic of single-colour samples rather than a finished picture.

Reconstructing a full-colour image therefore requires an interpolation process called demosaicing, which estimates the two missing colour values at every site from the neighbouring samples. The simplest approach, bilinear interpolation, averages nearby values, but it can produce pseudocolour where colours change sharply. More refined algorithms detect edges and interpolate along them rather than across them. Poor demosaicing produces false colour along edges and moiré patterns on fine, repetitive detail. Many sensors therefore include an optical low-pass filter that softens very fine structure before it reaches the photosites, at the cost of some sharpness. Because the outcome depends on both the sensor design and the algorithm, the same raw data can yield noticeably different images in different cameras or software.

Once the colour values are complete, the camera must correct for the light source and express the result in a standard colour space. White balance removes the colour cast produced by the illumination; in its simplest form it scales the red, green and blue channels so that neutral surfaces appear neutral. This correction is normally applied to the linear sensor data before the camera applies nonlinear tone and colour processing. The processed result is usually encoded in sRGB, a standard space that uses gamma encoding to match the response of typical displays. Since raw colour values depend on the particular sensor, software that handles images from many cameras usually converts them into a device-independent space such as CIE XYZ before comparing or editing them. Correcting white balance after the camera has already rendered an sRGB file is harder, because the nonlinear processing has already changed the relationship between the recorded values and the original scene colours.

Digital Processing Pipeline and File Formats

Once a CMOS or CCD sensor has converted light into voltages, those voltages must be turned into numbers. An analog-to-digital converter assigns each pixel's voltage to one of a fixed set of integer codes. The number of codes is set by bit depth. Camera sensors typically capture at 12-bit or 14-bit depth. A 12-bit file offers 4,096 tonal values per channel, while a 14-bit file offers 16,384. Bit depth governs tonal steps rather than detail: bit depth has nothing to do with resolution or the sharpness of an image. Finer steps reduce visible banding in smooth gradients and leave more room for edits such as lifting shadows.

Digitised data then passes through the image signal processor (ISP), a chain of fixed-function or programmable stages. The first step is usually black-level correction. A sensor produces a small signal even in total darkness, and black-level correction removes that offset. Other typical stages include dead-pixel correction, lens shading correction, white balance, demosaicing of the colour filter mosaic, noise reduction, colour correction, and gamma encoding. Order matters. One patented arrangement moves noise reduction ahead of black-level correction, because black-level correction applied first can clip black noise and cause nonlinearity in dark areas.

Gamma encoding converts linear light values into a nonlinear representation that matches how human vision responds to brightness. Nonlinear gamma encoding, such as the sRGB transfer, maps linear scene luminance into a perceptually more uniform space for standard displays. Sharpening follows, with sharpening boosting high-frequency components to improve perceived crispness. The processed result is then quantised and encoded to a file format, possibly with chroma subsampling and embedded metadata.

The two main output paths are RAW and processed files. RAW files store linear sensor data before demosaicing and white balance are applied. Because the data is not yet a finished picture, RAW processing software is needed to produce editable images. Most RAW captures keep the full sensor bit depth, which is typically 12 or 14 bits. A JPEG, by contrast, is an 8-bit format in that each colour channel uses 8 bits to describe the tonal value of each pixel.

Compression reduces file size, and it comes in two forms. Lossless compression, the kind used in general data compression and in lossless methods, restores the original data exactly but saves less space. Lossy compression discards information that the encoder judges less perceptible, such as fine texture and subtle colour transitions, in exchange for much smaller files. JPEG is the familiar lossy format, and its quality setting controls how much detail is sacrificed. Newer formats can shift this trade-off: HEIF files are typically no larger than JPEGs even though they are 10-bit files and JPEGs are 8-bit. Repeated lossy saving accumulates further loss, which is why editing work is usually kept in RAW or other high-precision formats until the final export.

Exposure Control: Shutter, Aperture, and Sensitivity

Exposure is the total light energy per unit area that reaches the sensor or film, which can be written as $H = E \times t$, where $E$ is the illuminance at the image plane and $t$ is the time it is admitted. Three controls set this quantity. Shutter speed sets $t$, aperture sets how much light passes through the lens (see optics), and sensitivity (ISO) sets how strongly the recording medium responds to a given amount of light. Because these settings are interdependent, a photographer can often reach the same brightness with different combinations. Changing the shutter speed also changes motion blur, and raising ISO amplifies signal and noise together, which is why the three settings are usually considered as a single "exposure triangle" rather than in isolation.

Shutter Mechanisms

The shutter determines how long light is admitted. In focal-plane designs, two curtains travel across the film or sensor, and at fast speeds the second curtain follows the first, so only a moving slit is open at any moment. Modern focal-plane cameras typically synchronise with flash only up to about 1/125 to 1/250 of a second, because a single flash burst must find the whole frame uncovered. Leaf shutters, by contrast, consist of blades in the lens that open and close like an iris, exposing the whole frame at once. This allows flash synchronisation at any speed up to the shutter's maximum, though at the fastest settings the blades may clip the tail of the flash pulse. Leaf shutters are also quieter, and they permit one shutter to serve interchangeable lenses in some medium-format and large-format systems, while focal-plane shutters remain the norm in small-format cameras because they reach very high top speeds more cheaply.

Electronic Shutters and Rolling Shutter

An electronic shutter controls exposure by resetting and reading out the sensor's pixels without any moving parts. This makes it silent and can allow very high speeds, and it suits video because no mechanism needs to cycle. The trade-off is the rolling shutter effect. Most CMOS sensors read their rows in sequence, so different parts of the image are recorded at slightly different times. A fast-moving subject can be recorded in different positions across its height, producing skewed or wavy shapes in propellers, vehicles, or panning shots. Global-shutter sensors avoid this by exposing all pixels at the same instant, though they have traditionally been more costly. Designers also use electronic correction in software to reduce the distortion, and many cameras offer a mechanical or electronic-front-curtain option so the photographer can choose between the two behaviours.

Metering and Automatic Exposure

Exposure metering measures the brightness of the scene, typically through the lens or with a dedicated sensor, and converts it into a recommended combination of shutter speed, aperture, and ISO. Common metering patterns weight the measurement toward the centre, average the whole frame, or evaluate small spot areas. Because a meter assumes that the scene reflects a typical average amount of light, bright snow or dark subjects can be rendered too dark or too light unless the photographer applies exposure compensation. Automatic modes use the metered value to set one or more parameters, such as aperture priority, where the photographer chooses the aperture and the camera chooses the shutter speed, or fully automatic programs that choose all settings. These modes depend on the reliability of the metering system, which is why they can misjudge high-contrast scenes.

Practical Trade-offs

Each setting has consequences beyond brightness. A short shutter time freezes motion but needs more light or higher sensitivity, and the added sensitivity raises image noise (see photography for the broader context of image capture). A small aperture increases depth of field but reduces the light reaching the sensor and can introduce diffraction. Modern sensors and computational methods, including the CMOS technology used in most current cameras, have made high sensitivity far more usable than in the film era, shifting the balance of these trade-offs in practice.

Focusing and Autofocus

Focusing adjusts the lens so that the image of a chosen subject falls exactly on the recording plane. In the thin-lens approximation, $\frac{1}{f} = \frac{1}{d_o} + \frac{1}{d_i}$, where f is the focal length, d_o the subject distance and d_i the image distance. The sensor or film stays at a fixed position, so changing focus means changing the effective image distance, usually by moving one or more lens elements along the optical axis. In manual focusing, a rotating ring drives this movement through a helical or cam mechanism. The photographer judges sharpness on a ground-glass screen, a split-image or microprism aid, or an electronic viewfinder. Large-format view cameras achieve the same result by moving the lens standard or the film holder on bellows.

Contrast-detection autofocus

Contrast-detection autofocus uses the image sensor as its own measuring device. The processor computes a sharpness value over a selected region, often based on the strength of high-spatial-frequency detail or local brightness differences. The lens motor then moves the focusing group while this value is tracked. A defocused image is blurred, so the value rises toward a single peak at best focus. The controller climbs toward that peak, overshoots, and reverses, which is why the lens can appear to hunt back and forth before settling. The method needs no extra hardware and is precise, which makes it common in mirrorless cameras and in live-view shooting. Each measurement requires a new frame, however, so it is relatively slow. It also struggles in low light and on low-contrast surfaces such as plain walls or clear sky.

Phase-detection autofocus

Phase-detection autofocus can determine the focus error from a single reading. Light arriving through one side of the lens aperture and light arriving through the opposite side form two images. When the subject is out of focus, these two images are displaced relative to each other; when the subject is in focus, they coincide. The sign of the displacement shows which way the lens must move, and its size indicates how far. In traditional single-lens reflex cameras, a beam splitter or secondary mirror directs light to a dedicated sensor with paired line arrays. Many modern sensors instead contain pixels that are split or masked so that each half sees only one part of the lens pupil, which allows phase detection across much of the frame. Because the system knows both direction and magnitude, it can drive the lens almost directly to focus. This makes the method fast and suitable for moving subjects. Its accuracy, however, depends on careful calibration of the optical paths.

Hybrid systems and subject tracking

Current systems usually combine both approaches. Phase detection provides a rapid, approximate lens movement, and contrast detection refines the final position. Subject tracking adds a further layer. Face, eye and body detectors, often built on convolutional neural networks, identify targets in each frame; object-detection methods then estimate where the target is moving, and the focus point follows it. This behaviour is known as continuous or servo autofocus. The overall process works as a feedback loop: the camera measures focus error, adjusts the lens, and measures again. Video autofocus must also move the lens smoothly, because abrupt focus changes are visible in footage, so lens motors are designed for quiet, gradual adjustment. These algorithms rely on the same sensor data and image processing pipeline used for the picture itself, and their reliability in crowded or cluttered scenes remains a practical limit.

Principal Camera Types

The large-format view camera is the oldest of the designs still in use. Its front standard holds the lens and its rear standard holds the film, joined by a light-tight bellows. A large format camera is nothing more than a place to put film, a place to put a lens, and bellows between the two. Because each standard can move independently through rise, fall, shift, tilt and swing, the photographer can control the plane of focus, depth of field and converging verticals. Monorail versions mount both standards on one rail and suit studio work, while folding field versions are lighter and popular with landscape photographers. Most view cameras take sheet film, and 4×5 inches is the most common size. The cost of this control is slow, deliberate operation and a tripod-bound workflow.

The single-lens reflex (SLR) camera solved the problem of framing: the viewfinder shows what the lens sees. A mirror set at an angle reflects the image upward to a focusing screen, and a pentaprism then corrects the reversed image so it appears upright. Innovations that transformed the SLR were the pentaprism eye-level viewfinder and the instant-return mirror—the mirror flipped briefly up during exposure, immediately returning to the viewfinding position. The first 35 mm SLR with a built-in pentaprism was the Contax S of 1949, and the design became the standard for professional and serious amateur cameras in the film era. When digital sensors replaced film, the same optical architecture was retained in digital SLRs (DSLRs), with a sensor in place of the film plane. The mirror box and pentaprism, however, add bulk and weight to the body.

Mirrorless cameras remove the reflex mirror and the optical viewfinder altogether. Light falls directly onto the sensor, and the image is shown on a rear screen or in an electronic viewfinder (EVF) that displays the live sensor output. Without the mirror box, the lens-to-sensor distance can be shorter, which makes bodies smaller and lighter and allows lenses from other systems to be adapted. The electronic viewfinder also previews exposure and other settings before the shutter is pressed, and it works with autofocus methods that use the sensor itself. Since the 2010s, mirrorless designs have taken a growing share of interchangeable-lens sales, and several major manufacturers have shifted their flagship development from DSLRs to mirrorless systems.

Several other designs occupy different points in the trade-off between portability, image quality and durability. Rangefinder cameras use a separate viewfinder whose optical path is coupled to the lens focus, so framing does not pass through the taking lens; they are quiet and compact, but they are awkward with long telephoto lenses. Compact or point-and-shoot cameras combine a fixed lens with a small sensor and automatic controls, and they were largely replaced in everyday use by smartphones. Medium-format cameras use film or sensors larger than the 35 mm format, which gives greater detail and tonal range at the cost of size, price and speed. Action cameras are small, wide-angle and often waterproof or shock-resistant, and they are designed for mounting on helmets, vehicles or other moving objects rather than for handheld composition.

These categories are not rigid, and boundaries have blurred as sensors, lenses and smartphone imaging have improved. A large-format view camera, a camera obscura and a modern mirrorless body all still rely on the same basic optical principles of forming an image from light passing through a lens, but they differ in how they frame, focus and record that image.

Motion-Picture and Video Cameras

Motion-picture cameras expose a rapid series of still images, and the viewer's perception of smooth movement depends on how fast those frames follow one another. The traditional explanation credits persistence of vision, but that account is disputed. Critics describe the idea that motion perception comes from the eye alone as a myth, and Max Wertheimer's work in 1912 is cited as having undermined it. Many researchers now prefer to attribute the effect to perceptual illusions such as the phi phenomenon and beta movement.

Frame rate and flicker are separate matters. Early exhibitors learned that a slow frame rate can be made acceptable by interrupting each image. The Lumière brothers flashed each frame several times, using 16 frames per second with three showings per frame, which gave a 48 Hz flicker rate that avoided visible flicker while saving film. In film systems the frame must be moved between exposures, so a shutter has to hide that movement. Film therefore needs at least one interruption per frame.

Film cinematography depended on a precise mechanical transport. During each cycle the shutter opens to expose the frame while registration pins hold the film steady, and then the shutter closes while a claw pulls the next frame into the gate (the shutter opens, exposing the light-sensitive film stock in the film gate for a split second, while registration pins hold the film still). The film stops completely for every exposure, so each frame is a separate still photograph (each exposure is a single still photograph, or frame). Engineers also had to reduce the wear that the claw caused on the perforations and the vibration of cam-driven movements. Film stock itself changed as well: Kodak moved from acetate to a polyester base that could be pulled through the camera faster and was more dimensionally stable. Standard speeds became fixed by convention. Theatrical film is commonly shot at 24 frames per second, and 30 frames per second is the broadcast standard in North America and Japan, while 25 frames per second is standard in much of the rest of the world. Cinematographers also follow a convention of setting the shutter to a 180-degree opening, which keeps the motion blur familiar to audiences.

Digital cinema cameras and camcorders replace the film gate with electronic image sensors, and the recording format now defines the characteristics that film stock once fixed. Among these are frame rate, spatial resolution, colour depth, and the compression applied to the stream. Colour depth is measured in bits per channel, and the choice of compression, a form of data compression, determines both file size and how much detail survives. CMOS sensors read out pixels individually, which makes higher frame rates practical (in CMOS cameras each pixel can be read out individually), and this capability has spread to consumer devices, where mobile phones and consumer cameras provide frame rates of 120-240 fps at relatively low resolutions and with a rolling shutter. Rolling-shutter readout can skew fast-moving subjects, a limitation that becomes more serious as speed increases.

High-speed cameras form a specialised class defined by their capture rate. A high-speed camera is a device capable of capturing moving images with exposures of less than 1/1000 second, or frame rates in excess of 250 frames per second. Modern electronic models typically record over 1000 frames per second, which are then played back slowly to study the motion of transient phenomena. The method is useful because the human eye cannot follow very fast events. For example, a hummingbird flaps its wings up to 80 times per second, too fast to distinguish flaps with the human eye, yet a high-speed camera slows the wing rate to an observable speed. Applications include observation of collision phenomena and shockwaves, and destructive observation in material testing. Because such recordings produce large volumes of data, some systems keep recording continuously into a buffer and stop once a trigger detects the event (by utilizing an electronic trigger the high speed camera is continuously recording and replacing the oldest images with newer images, and once the event occurs, a machine vision sensor stops recording).

Camera Phones and Computational Photography

Camera phones combined a small CMOS image sensor with one or more compact lenses in a device that was already carried everywhere. The sensor and lens are limited by the phone's thickness. Their focal lengths are short and their apertures are small, so they cannot reproduce the optical behaviour of a large-format or DSLR lens. Over time the smartphone became the most widely used camera in the world, and much of the image quality it delivers comes from software rather than optics. This shift is usually described as computational photography, a term for techniques that use digital processing to form or improve an image.

Most phones now have several rear modules, typically a wide-angle lens, a telephoto lens and an ultra-wide lens. Each module covers a different angle of view, so the phone can switch between them as the user zooms, or blend data from neighbouring modules to produce intermediate focal lengths. Where two modules view the same scene from slightly different positions, the difference between their images (stereo disparity) is also used to estimate the distance of objects, which feeds later processing.

High dynamic range (HDR) imaging is the most common application. A single sensor exposure cannot capture both bright skies and deep shadows, because the sensor's dynamic range is limited. The phone therefore captures several frames at different exposure levels in quick succession, aligns them to compensate for hand movement, and merges them. Some implementations combine more than ten frames. The result keeps detail in highlights and shadows, and modern implementations, which are automatic, look far more natural than the over-processed images of earlier years. Night modes apply a related principle at much lower light levels. The phone captures a burst of longer exposures, uses software to correct for the movement of the hand, and averages the frames. Averaging suppresses random sensor noise, so faint detail becomes visible. Google calls its version Night Sight and Apple calls its version Night Mode, and both can take a few seconds of capture time.

Portrait mode simulates the shallow depth of field, in which the subject is sharp and the background is softly blurred. A physical shallow depth of field requires a long focal length or a wide aperture, which a phone cannot provide. Instead, the device estimates a depth map of the scene, either from two cameras, a dedicated depth sensor, or a machine-learning model that segments the subject from the background. It then applies a graduated blur to the background, with the blur increasing with distance from the subject. The method is imperfect: fine structures such as hair, fabric edges and thin objects are difficult to segment, and errors in the depth map can create halos or blurred regions that do not match the scene. These limitations are a known subject of ongoing development.

These features require substantial processing power, and modern phones include specialised image signal processors and neural-processing hardware for the purpose. Tasks such as face detection, segmentation, multi-frame merging and noise reduction are carried out while the picture is being taken, which allows the results to appear almost immediately in the viewfinder. The consequence is that a phone's photographic quality depends heavily on its software and its image-processing design as well as on its hardware, and manufacturers differentiate their products partly through these algorithms.

Computational photography has also changed what users expect from a camera. Features that once required manual bracketing, long-exposure technique or large lenses are now automatic. Critics note that heavy processing can smooth away texture, alter colours and produce results that differ from the scene as it appeared, which raises questions about how much of a smartphone photograph represents the physical scene. This tension between faithful recording and enhancement is discussed further in the sections on image reconstruction and on the social and ethical dimensions of photography.

Scientific, Industrial, and Specialised Cameras

Astronomical cameras use the same kinds of CMOS and CCD sensors as consumer devices, but they are cooled to suppress dark current, the signal produced by heat rather than light. Dark current rises quickly with temperature, so detectors are often held well below ambient, usually with thermoelectric (Peltier) coolers in modern instruments. A faint galaxy may deliver only a few photons per pixel, so observers record many long exposures and combine them. Electronic detectors replaced photographic plates on large telescopes during the 1980s and 1990s, because they are far more sensitive and their output is linear and easy to calibrate.

Thermal (infrared) cameras form images from the radiation that objects emit according to their temperature, as described by blackbody radiation. They need no external light and can see through darkness or some smoke. Most handheld units use uncooled microbolometer arrays, in which each pixel's electrical resistance changes as it absorbs long-wavelength infrared, typically in the 8–14 µm band of the electromagnetic spectrum. Cooled detectors made from indium antimonide or mercury cadmium telluride offer higher sensitivity and appear in military, scientific, and high-end inspection equipment. In building inspection, thermal images reveal missing insulation, air leaks, and moisture. Firefighters use them to locate people and hot spots through smoke. In medicine, they are used to map skin temperature and surface blood flow, although thermal images are generally treated as supporting information rather than stand-alone diagnoses.

Microscope cameras attach to the optical path of a microscope and typically use monochrome, cooled, or scientific CMOS (sCMOS) sensors to capture faint fluorescence from stained specimens. Endoscopes take a different approach: a miniature sensor sits at the tip of a rigid or flexible tube, a design often called chip-on-tip, which lets clinicians inspect body cavities. Industrial borescopes use the same principle to examine the interior of jet engines, pipes, and machinery. Scanners and industrial line-scan cameras build images by sweeping a linear sensor across a document, web of material, or production line, which suits the inspection of printed circuit boards, textiles, and food. In all these fields, the illumination is often as important as the sensor. Ring lights, backlights, ultraviolet or infrared sources, and structured-light patterns determine which features the camera can capture with useful contrast.

Satellite and aerial cameras observe the Earth from orbit, aircraft, and drones as part of remote sensing. Unlike a consumer camera, which records three broad colour channels, many remote-sensing instruments record multispectral bands that extend into the near-infrared, shortwave infrared, and thermal infrared, wavelengths the eye cannot see. The Landsat programme, whose first satellite launched in 1972, has supplied such multiband imagery for more than five decades. Analysts combine bands into derived indices. The normalised difference vegetation index, for example, compares near-infrared and red reflectance to estimate plant health, while thermal bands help track surface temperature for crop water stress and urban heat islands. Many spaceborne imagers use pushbroom designs, in which a line of detectors builds the image as the platform moves along its orbit. Hyperspectral instruments extend this idea to hundreds of narrow spectral bands, which makes it possible to distinguish materials with similar broadband colour.

Image Quality, Measurement, and Physical Limits

Sharpness is often described by pixel count, but engineers also use a contrast-based measure. The modulation transfer function (MTF) compares how much contrast an imaging system keeps relative to the original scene, and it does so across a range of fine-to-coarse patterns. MTF is the ratio of relative image contrast divided by relative object contrast. Perceived sharpness is closely tied to the spatial frequency at which the MTF has fallen to half. A complete camera's response is the product of the MTFs of its parts, so a high-megapixel sensor paired with a weak lens does not deliver its nominal detail.

Diffraction sets a limit that no lens design can remove. Light passing through a circular aperture spreads into a central bright spot, the Airy disk, whose diameter is roughly 2.44 times the wavelength multiplied by the f-number. Taking 550 nm as the wavelength gives an Airy disk of about 6.7 µm for an f-number of 1 in that calculation; at f/16 the same formula gives about 21 µm. This limit depends on the f-number and the wavelength, not on how many megapixels the sensor has. Pixel size decides when the blur becomes visible. Larger pixels hide diffraction until smaller apertures, so a 12-megapixel camera can be stopped down further before softening appears than a 36-megapixel one. This creates the familiar trade-off: a smaller aperture increases depth of field, but beyond an optimum it reduces sharpness.

Noise has a physical floor. Light arrives as discrete photons, and their arrival counts fluctuate randomly. The standard deviation of such fluctuations equals the square root of the average count. In practice, this means that an average of 10,000 photons fluctuates by about 100, so the relative noise is larger in dark areas where fewer photons arrive. Photon shot noise follows Poisson statistics, and shot noise typically dominates the sensor noise for signal levels above 10 photons. Electronic read noise adds to this, and the two together set the practical noise level. Raising sensitivity mainly applies amplification after the signal has been captured, so the noise already present is amplified along with the image. Even with a perfect sensor adding no noise, dynamic range would eventually be limited by photon shot noise.

Dynamic range describes how much tonal range a sensor can record in one exposure. It is the ratio between the largest signal the sensor can hold before saturating and the smallest signal that rises above the noise. The upper end of this range is set by the full well capacity, the maximum charge a photodiode can store. Dynamic range is conventionally expressed in decibels, using 20 times the base-10 logarithm of the ratio. For example, one measured sensor was reported at 54 dB, which corresponds to a ratio of roughly 500 to 1. Because each photographic stop is a doubling of light, 6 dB corresponds to about one stop, so a sensor's dB figure can be read as a stop count.

These measures are interdependent. Resolution figures ignore contrast, MTF ignores noise, and dynamic range ignores fine detail. A meaningful assessment of image quality therefore requires all of them, evaluated under controlled test conditions with the lens, sensor, and processing treated as a single system.

Computer Vision, Artificial Intelligence, and Machine Perception

Cameras are the main data source for computer vision, a field that converts pixel arrays into descriptions of objects, surfaces, and events. Common tasks include image classification, which assigns a label to a whole image; object detection, which locates and labels several objects with bounding boxes; semantic segmentation, which assigns a class to every pixel; and tracking, which follows the identity of an object across successive frames. Before about 2012, these systems relied on hand-designed features such as SIFT keypoints and HOG descriptors, passed to classifiers such as support vector machines. Progress on benchmarks was steady but incremental.

The field changed with deep learning. Convolutional neural networks learn their own filters from data rather than relying on engineered features. In 2012, the AlexNet entry in the ImageNet Large Scale Visual Recognition Challenge, trained on roughly 1.2 million labelled photographs, reduced the top-5 error rate to about 15 percent, well ahead of the next-best entry at about 26 percent. Subsequent architectures such as VGG, GoogLeNet and ResNet extended the approach, and ResNet's residual connections made very deep networks trainable. These models depend on machine learning from large labelled collections and on GPU hardware for training. Transformer-based vision models have since joined convolutional ones. Performance is closely tied to the training data: a model tends to work well on images resembling its training set and can degrade under unfamiliar lighting, viewpoints, or populations.

Mobile platforms such as robots, drones, and autonomous vehicles usually combine cameras with radar, lidar, inertial sensors, and satellite positioning, and the camera contributes geometric information. Stereo depth estimation uses two cameras separated by a known baseline $B$. For focal length $f$ and disparity $d$ between matched points, the depth is $Z = \frac{fB}{d}$, so depth precision depends on calibration accuracy and disparity resolution. Calibration estimates intrinsic parameters such as focal length, principal point, and lens distortion, usually with a target of known pattern, and extrinsic parameters that relate the camera's pose to other sensors. Visual navigation builds on these estimates through simultaneous localization and mapping (SLAM) and visual odometry, in which the estimated pose is fed back into the control loop of the vehicle or robot.

Face and biometric recognition is a specialised application of these methods. A system first detects faces, then maps each face to a numerical feature vector and compares it with stored templates. Verification compares one face with one claimed identity, while identification searches a large gallery. Performance is reported with false match and false non-match rates, which depend on threshold choices and image quality. In its 2019 evaluation of face recognition algorithms, the US National Institute of Standards and Technology found that many algorithms showed differences in error rates across demographic groups, and that the size of these differences varied widely between algorithms. Legal treatment also varies. Under the EU's General Data Protection Regulation, biometric data processed for the purpose of unique identification is a special category of personal data. The EU Artificial Intelligence Act, adopted in 2024, restricts real-time remote biometric identification in publicly accessible spaces by law enforcement, subject to narrow exceptions. Some municipalities have banned government use of the technology. These issues connect the technical questions of accuracy and calibration with wider debates on data privacy and statistical fairness.

Industry, Economics, and the Decline of Film

Over roughly two decades after 2000, the centre of the photographic industry moved from companies that sold film, chemicals and printing to companies that make electronics, sensors and software. Film had produced recurring revenue for each exposure, and digital capture removed that stream. Buyers began paying for devices, storage, editing software and online sharing instead. Demand for dedicated cameras also fell as smartphones took over most casual photography.

Kodak illustrates the transition. Its business rested on cheap cameras sold through retailers, which led to profits from film, paper and processing. Kodak sold cheap cameras through a network of retailers. It earned money with expensive complements, namely photo films, paper, and a printing service. Kodak engineers built a working digital camera in 1975, but management regarded digital images as a threat to film and did not pursue the product commercially. Kodak filed for Chapter 11 bankruptcy protection in January 2012, and in September 2013 it emerged from bankruptcy after shedding large legacy liabilities and exiting several businesses. Kodak's experience is often used as a case study in innovation and technological change, and in the difficulty incumbents face when a new technology undermines their core revenue, a dynamic sometimes described as creative destruction.

Patents also shaped the industry's economics. Kodak pursued patent enforcement to offset falling sales, and in 2010 it earned $838 million, including a settlement with LG Corporation. Licensing revenue of this kind was not a substitute for product sales, but it shows how intellectual property became a source of income and a battleground for firms competing in digital imaging.

The volume decline is large and well documented, although the exact percentages vary by source. According to historical CIPA data, camera shipments peaked during the 2000 to 2010 digital boom, when more than 120 million units were sold. By 2023, the shipments of fixed-lens cameras had collapsed: CIPA members shipped just 1.7 million cameras with a built-in lens, down from almost 109 million in 2010. Over a shorter span, worldwide camera shipments dropped by 84% between 2010 and 2018, according to CIPA. The collapse was concentrated in the compact segment, while interchangeable-lens cameras held up far better.

The market that remains is smaller in units but more polarised. In 2025, DJI shipped more than 10 million handheld units, surpassing that milestone for the first time, while all CIPA member companies combined shipped 9.4 million cameras, including mirrorless cameras, DSLRs and compact cameras, of which 6.3 million were mirrorless. In the first half of 2026, handheld smart camera shipments rose by close to 50%, while CIPA's interchangeable-lens shipments fell by about 5%. Traditional manufacturers have therefore kept revenue per unit up even as volumes shrink, with shipment volumes declining slightly, but the value of those shipments, and therefore profits, increasing.

Production of the underlying components has also become concentrated. Image sensors require fabrication facilities that cost large sums to build and operate, precision lens manufacturing depends on specialised glass and tooling, and competitive image processing relies on large software and algorithm teams. These capital and know-how requirements limit the number of firms that can compete at the high end, so the industry's structure now depends on a small group of suppliers rather than on the many film and camera makers of the twentieth century.

Social, Legal, and Ethical Dimensions

Cameras have changed how societies record events, and that change has raised questions about truthfulness, privacy, authorship and evidence. Each area has developed its own rules, often years after the technology became common.

Credibility in photojournalism and documentary work

Photographs carry strong persuasive force, which makes them easy to misuse. Professional codes address this directly. The U.S. National Press Photographers Association asks visual journalists to resist staged photo opportunities and not to alter images in ways that mislead viewers. The Canadian Association of Journalists says altered images should be labelled as photo illustrations. Many codes allow minor technical adjustments, such as cropping that leaves context intact or conversion to grayscale, but treat changes that alter the factual content as unacceptable. Digital editing made such alterations far easier than in the film era. In 2015, World Press Photo withdrew a first-place prize after finding that the photographer had misrepresented where a picture was taken, and the organisation later disqualified a large share of finalists under new post-production rules.

Privacy and personal data

Cameras now operate in streets, shops, workplaces and nearly every pocket, so the question of who may capture and keep identifiable images has become a legal matter. In the European Union, the data protection framework of the General Data Protection Regulation, in force since 25 May 2018, treats an image as personal data when a person can be identified from it, whether through a visible face or distinctive features. Facial images processed with technical methods to confirm or establish a unique identity fall into a "special category" with stricter conditions, such as explicit consent or a substantial public interest basis. Anonymous crowd scenes in which no individual can be distinguished generally fall outside the rule. The EU's AI Act adds transparency duties for AI-generated content, including deepfakes, and these became enforceable in August 2026. Other jurisdictions take different approaches, and the balance between public-safety uses, such as surveillance and live facial recognition, and individual privacy remains disputed.

Copyright and authorship

Photographs are generally protected as works of authorship under copyright law, which grants the creator exclusive rights over reproduction, distribution and public display. The Berne Convention, adopted in 1886 and revised several times since, established that protection arises when a work is fixed in tangible form, without registration, and that authors from one member state receive equivalent protection in the others. The Convention sets a minimum term for photographs of 25 years from the year of creation, a shorter period than the minimum set for most literary and artistic works, although many countries grant longer terms. Technological change keeps reopening questions: how to assign rights in images altered by software, who holds rights in material produced by automated or AI systems, and how ownership passes when photographs are commissioned or made by employees. Answers differ by country and by contract, and they are still being worked out in courts and legislatures. These questions connect to wider intellectual-property debates.

Evidence and the integrity of records

Photographs and video are used as evidence in courts, insurance claims and security systems. Their weight depends on whether they can be shown to be unaltered. Investigators therefore document the chain of custody, from capture through storage and transfer, and examine metadata that records the device, settings and edit history. Because digital files can be changed without leaving obvious traces, the reliability of an image as a record depends less on the picture itself than on the controls around its creation and storage. As generative artificial-intelligence tools make convincing fabrications easier to produce, these verification practices have become a central concern for courts, news organisations and insurers alike.

References

1. Pinhole Camera: Working Principle, Diagram & Uses Explainedvedantu.com
2. The Pin hole Camera (docx) - CliffsNotescliffsnotes.com
3. [FREE] Explain the principle of a pinhole camera and state three characteristics of the image formed. Provide a - brainly.combrainly.com
4. Anti-biofouling Lensless Camera System with Deep Learning based Image Reconstructionarxiv.org
5. Mid-infrared nonlinear pinhole imagingarxiv.org
6. Thin Lens Equation Calculatorpearson.com
7. Thin Lens Equation Interactive Calculatorfirgelliauto.com
8. Method for electronically configurable variable aperture and gratingimage-ppubs.uspto.gov
9. History of the cameraen.wikipedia.org
10. Pinhole cameraen.wikipedia.org
11. Camera Obscuraebsco.com
12. CAMERA OBSCURA TIMELINE 1.5: AL-HAYTHAM (ALHAZEN)earlymovingimagetimeline.wordpress.com
13. graphsearch.epfl.chgraphsearch.epfl.ch
14. The Universe Through a Pinhole: Ibn al-Haythamoerproject.com
15. Camera Obscuraitakehistory.com
16. History of photographyen.wikipedia.org
17. Camera Obscura: The Ancient Lens of Culture and Tomorrow’s Visionmedium.com
18. Camera obscura Painting techniques Oil Canvas Famous painters Stylesreviewpainting.com
19. Camera obscurasimple.wikipedia.org
20. Camera obscuraen.wikipedia.org
21. Book of Opticsislam.fandom.com
22. Who Was Ibn al-Haytham - 1001 Inventions1001inventions.com
23. The First Photograph in Historyamusingplanet.com
24. Bitumen of Judeaen.wikipedia.org
25. History of Informationhistoryofinformation.com
26. Louis Daguerreen.wikipedia.org
27. Conservation and restoration of photographsen.wikipedia.org
28. Daguerreotypedesign-encyclopedia.com
29. Photographsmanual.museum.wa.gov.au
30. William Henry Fox Talbotbritannica.com
31. Calotype Paper Negatives - Tim Layton Fine Arttimlaytonfineart.com
32. Photography and Medicine: The Presidential Address, Delivered on 9th October, 1935, at the Opening of the Sixty-Third Session of the Bristol Medico-Chirurgical Societyncbi.nlm.nih.gov
33. After Niépce and Daguerre - Nicéphore Niépce's House Museumbicentennial-of-photography.org
34. Nicephore Niepcebritannica.com
35. Kodak Film Camera Is Patented and Trademarkedebsco.com
36. George Eastman patents Kodak camerahistory.com
37. Original Kodak Camera, Serial No. 540americanhistory.si.edu
38. Kodak camerabritannica.com
39. From the Camera Obscura to the Revolutionary Kodakeastman.org
40. Oskar Barnack - Camera-wiki.org - The free camera encyclopediacamera-wiki.org
41. The American Society of Cinematographerstheasc.com
42. Edwin Land and the Birth of Instant Photographyoptica.org
43. 75 years of instant photos, thanks to inventor Edwin Land’s Polaroid cameratheconversation.com
44. The first Polaroid camera is soldhistory.com
45. Edwin Land and Instant Photographyacs.org
46. Edwin Land: The Visionary Behind Polaroid and Instant Photographyotislibrarynorwich.org
47. Original Kodak Camera, Serial No. 2443si.edu
48. History of Photography: Leica, Barnack & the Electronic Flashifolor.ch
49. Ernst Leitz GmbHen.wikipedia.org
50. Leica copiesen.wikipedia.org
51. NIHF Inductee Willard Boyle Invented the Charge Coupled Deviceinvent.org
52. Charge-coupled devicenokia.com
53. CCD vs CMOS Image Sensors: A Comprehensive Guide for Engineering Professionalswevolver.com
54. CMOS image sensor with non-contact structureimage-ppubs.uspto.gov
55. CMOS image sensor readout employing in-pixel transistor current sensingimage-ppubs.uspto.gov
56. CCD vs. CMOS Sensors: Key Differences Explainedva-imaging.com
57. Milestones:Handheld Digital Camera, 1975 - Engineering and Technology History Wikiethw.org
58. In 1975, young Kodak engineer Steven Sasson built the first self-contained digital camera, a 3.6-kilogram device that took 23 seconds to record a 0.01-megapixel black-and-white image onto cassette tape—and Sasson later summed up management’s initial reaction at the film giant as: “That’s cute—but don’t tell anyone about it.” - Silicon Canalssiliconcanals.com
59. Apple QuickTakeen.wikipedia.org
60. Milestones:Charge-Coupled Device, 1969 - Engineering and Technology History Wikiethw.org
61. The ultimate image sensor guide: EMCCD vs. SCMOS vs. CMOS vs. CCD - e-con Systemse-consystems.com
62. Kodak engineer's prototype began the digital revolutiondeseret.com
63. Say Cheese! Meet the Man Who Invented the Digital Camera - National Science and Technology Medals Foundationnationalmedals.org
64. Adaptive Residual Interpolation for Color and Multispectral Image Demosaickingncbi.nlm.nih.gov
65. File:Tartan Ribbon.jpgen.wikipedia.org
66. Color Photography Turns 150 Years Old Todaypopphoto.com
67. Bayer_Interpolationmicrochip.com
68. Bayer Pattern - an overviewsciencedirect.com
69. Bayer filteren.wikipedia.org
70. Inheriting Bayer's Legacy-Joint Remosaicing and Denoising for Quad Bayer Image Sensorarxiv.org
71. Color balanceen.wikipedia.org
72. Perception-Inspired Color Space Design for Photo White Balance Editingarxiv.org
73. Beyond raw-RGB and sRGB: Advocating Access to a Colorimetric Image Statekaraimer.github.io
74. When Color Constancy Goes Wrong: Correcting Improperly White-Balanced Imagesopenaccess.thecvf.com
75. s11760 022 02230 2link.springer.com
76. Bit Depthphotographyicon.com
77. The Pitfall of Darkening a Sky, Part 2adorama.com
78. Black Level Correction in Image Sensors: What It Is and Why It Matters - e-con Systemse-consystems.com
79. Method and apparatus for post noise reduction black level correctionimage-ppubs.uspto.gov
80. Task-Aware Image Signal Processor for Advanced Visual Perceptionarxiv.org
81. 8-Bit vs 16-Bit Photos: Why Bit Depth Matters More Than You Thinkframeandfocal.com
82. Bits and bit depth explained - Life after Photoshoplifeafterphotoshop.com
83. Image File Formats: RAW, JPEG & more - Canon Georgiacanon.ge
84. GitHub - cruxopen/openISP: Image Signal Processor · GitHubgithub.com
85. Leaf vs Focal Plane Shutter: Key Differencesfstoppers.com
86. A Guide to Camera Shutter Designs: Focal-Plane, Leaf, and the Restpetapixel.com
87. Shutter (photography)en.wikipedia.org
88. Focal-plane shutteren.wikipedia.org
89. Rolling shutteren.wikipedia.org
90. Electronic shutter vs mechanical shutter - Canon Europecanon-europe.com
91. The rolling shutter effect explained - Videomakervideomaker.com
92. Choosing a 4x5 Camera and Lenses — Alex Burke Photographyalexburkephoto.com
93. History of the single-lens reflex cameraen.wikipedia.org
94. View cameraen.wikipedia.org
95. An Introduction to Large Format Photographyphotography.tutsplus.com
96. A Beginner's Guide to Large Formatjollinger.com
97. Prakticaen.wikipedia.org
98. Single-lens Reflex Camera,Single-lens Reflex Camera inventorsedubilla.com
99. Digital single-lens reflex cameraen.wikipedia.org
100. A Brief History of the Mirrorless Camera Systemlightstalking.com
101. Does a mirrorless camera have a pentaprism » Camera Reviewsgoprocamerasreview.com
102. Motion Picture Persistence of Vision: "How You See It" 1936 Chevrolet 8min — Steemitsteemit.com
103. Motion picture systemimage-ppubs.uspto.gov
104. Intermittent mechanismen.wikipedia.org
105. Motion-picture camerabritannica.com
106. Film frameen.wikipedia.org
107. A trillion frames per second: the techniques and applications of light-in-flight photographyarxiv.org
108. Things to consider when buying a high-speed cameraix-cameras.com
109. High-speed cameraen.wikipedia.org
110. High Speed Cameras Selection Guide: Types, Features, Applicationsglobalspec.com
111. What is High-Speed Video Camera? : SHIMADZU (Shimadzu Corporation)shimadzu.com
112. High-speed photographyen.wikipedia.org
113. AI-Enhanced Mobile Photography: Trends and Tips for Capturing Stunning Shots on Your Smartphoneproedu.com
114. How to take better photos with your smartphone, thanks to computational photographytheconversation.com
115. Computational Photography Explained: How Phones Outsmart Traditional Camerasmedium.com
116. The Impact of Computational Photography on Modern Smartphone Camerasboltonelectronics.com
117. Computational photographytapsmart.com
118. How Computational Photography Improves Smartphone Photoshowtogeek.com
119. Understanding Computational Photography in Smartphones - PictureCorrectpicturecorrect.com
120. Enhanced image capture sharpeningimage-ppubs.uspto.gov
121. Towards Single-Lens Controllable Depth-of-Field Imaging via Depth-Aware Point Spread Functionsarxiv.org
122. Noise, Dynamic Range and Bit Depth in Digital SLRsphotonstophotos.net
123. Imaging array with improved dynamic rangeimage-ppubs.uspto.gov
124. How numbers can mislead: a closer look at dynamic range, Part 2: DPReviewdpreview.com
125. Systems and methods for providing low-noise readout of an optical sensorimage-ppubs.uspto.gov
126. Diffraction Limited Photography: Pixel Size, Aperture and Airy Diskscambridgeincolour.com
127. What Is Lens Diffraction?photographylife.com
128. Adjustable apodized lens apertureimage-ppubs.uspto.gov
129. Understanding resolution and MTFnormankoren.com
130. Sharpnessimatest.com
131. Automatic image sharpening in an electronic imaging systemimage-ppubs.uspto.gov
132. High performance photosensorimage-ppubs.uspto.gov
133. Circle of Confusion, the Airy Disk and Diffraction (Podcast 594)martinbaileyphotography.com
134. Why Kodak Failedthevalueengineers.nl
135. Kodaken.wikipedia.org
136. Kodak: A journey from creating memories to bankruptcybusinesshistory.domain-b.com
137. Cameras Statistics By Usage, Type, Market Size and Facts (2025)electroiq.com
138. Digital cameras: a dying breed after smartphones wipe out decades of industry growth - Photo Rumorsphotorumors.com
139. How smartphones devastated the camera industryweforum.org
140. DJI sells more mirrorless digital cameras than all traditional camera makers combined - Notebookcheck Newsnotebookcheck.net
141. DJI and Insta360 Shipped More Cameras Than the Mirrorless Marketdiyphotography.net
142. Privacy-preserving Object Detectionarxiv.org
143. A First Attempt at Regulating Biometric Data in the ...ainowinstitute.org
144. Protecting Persona Biometric Data: The Case of Facial Privacyarxiv.org
145. GDPR for Images: Compliance Overview for Visual Data Protection - GDPR Localgdprlocal.com
146. Processing biometric data? Be careful, under the GDPRiapp.org
147. Photojournalism and Ethicspsd-dude.com
148. Photojournalism ethicscjf-fjc.ca
149. Code of Ethics for Visual Journalistsnppa.org
150. Code of Ethicsfocuspointgazette.com
151. World Press Photo Addresses Ethics of Digital Photo Manipulationadorama.com
152. Berne Conventionen.wikipedia.org
153. Berne Conventionbritannica.com
154. Berne Convention for the Protection of Literary and Artistic Worksitlaw.fandom.com
155. Berne Conventiontheipmatters.com