Every phone camera, CCTV camera, medical scanner and self-driving car depends on an image sensor: a chip that turns a pattern of light into a pattern of electrical charge and then into numbers. Once the image is in numbers, its colours must be measured and reproduced faithfully on screens and prints. This module of Physics for Emerging Fields covers the construction and working of the two main imaging sensors, CCD and CMOS, how monochrome and colour images are formed, and the science of colour: the chromaticity diagram, chromaticity coordinates, colour measurement and colour matching. It maps to course outcome CO2 (measure chromaticity and illustrate the colour matching concept), and it supports the lab experiments on CMOS sensors, colour sensors and RGB LED chromaticity.
Summary
Imaging sensors: An imaging sensor is a two-dimensional array of tiny light-sensitive elements called pixels (picture elements). A lens forms an optical image on the array. Each pixel absorbs photons, and by the photoelectric effect in silicon each absorbed photon (with energy above 1.12 eV, so wavelength below about 1.1 μm) frees an electron. The number of electrons collected during the exposure (integration) time is proportional to the light intensity at that pixel multiplied by the exposure time. The charge of every pixel is then read out, converted to a voltage, and digitised by an analog-to-digital converter (ADC). The two main technologies, CCD and CMOS, differ mainly in how this charge is read out.
CCD construction: A charge-coupled device (CCD), invented in 1969 by Willard Boyle and George Smith at Bell Labs (Nobel Prize 2009), is an array of closely spaced MOS capacitors built on a p-type silicon substrate covered with a thin insulating oxide (SiO2), on which transparent polysilicon gate electrodes are laid. Each pixel has typically three (or four) gates. Channel stops (heavily doped strips) separate the columns. At the end of the columns there is a horizontal serial (readout) register, and at its end a single output node with a charge-to-voltage amplifier.
CCD working: A positive voltage on a gate repels holes from the silicon below it and forms a potential well for electrons. During exposure, photo-generated electrons collect in the well under each pixel; the amount of charge is proportional to the light falling there. After exposure the gate voltages are clocked in sequence (three-phase clocking): the well under the next gate is made deeper while the present one is made shallower, so each charge packet moves one step along the column, like buckets passed along a line (the bucket brigade). Each row is shifted in turn into the horizontal register, which then shifts the packets one by one to the output amplifier, where each packet is converted into a voltage. The transfer is almost perfect (charge transfer efficiency above 0.99999 per step), so CCDs give very uniform, low-noise images. Types: full-frame CCD (needs a mechanical shutter), frame-transfer CCD (a covered storage area receives the whole image quickly) and interline-transfer CCD (a masked vertical register next to each column of photodiodes, allowing an electronic shutter).
CMOS sensor construction and working: A CMOS image sensor, also called an active pixel sensor (APS), is made with the same CMOS process used for microprocessors. Each pixel has a photodiode and its own small circuit, usually three or four transistors: a reset transistor that empties the photodiode before exposure, a transfer gate (in the 4T pixel) that moves the charge to a floating diffusion node, a source follower that converts the charge to a voltage and buffers it, and a row-select transistor that connects the pixel to its column line. The chip also contains row decoders, column amplifiers, ADCs (often one per column) and timing logic. Working: the photodiode is reset, light generates electrons during the exposure time, the charge is converted to a voltage inside the pixel, and rows are selected one at a time so that each column line carries the pixel voltage to the column ADC. Since every pixel can be addressed by its row and column, a window of interest can be read directly (random access). Most CMOS sensors expose and read the rows one after another (rolling shutter), which can distort fast-moving objects; global-shutter CMOS sensors expose all pixels together.
CCD versus CMOS: CCDs have excellent uniformity, high fill factor and low noise, but they need several high clock voltages, consume more power, are slower to read and need separate chips for the ADC and processing. CMOS sensors consume much less power, operate at a single low voltage, read out very fast, allow random access and put the ADC and image processing on the same chip, so they are cheaper and smaller. Early CMOS sensors were noisier, but modern designs (pinned photodiodes, correlated double sampling, backside illumination, microlenses) have closed the gap, and CMOS now dominates phones, cameras and automotive vision, while CCDs remain in some scientific and astronomical cameras.
Image formation, monochrome: A lens focuses the scene on the sensor. Each pixel collects a charge proportional to the light energy reaching it, irrespective of colour. The ADC converts each pixel signal into a number called the grey level. With 8-bit conversion there are \(2^8 = 256\) grey levels, from 0 (black) to 255 (white). The image is thus a matrix of grey levels, f(x, y), with as many rows and columns as the sensor has pixels. Spatial resolution depends on the number of pixels (and the lens), and intensity resolution depends on the number of bits per pixel.
Image formation, colour: A silicon pixel cannot tell colours apart, so colour is obtained by filtering. In most cameras a colour filter array called the Bayer filter (proposed by Bryce Bayer of Kodak in 1976) is placed over the pixels in a repeating 2 × 2 pattern of one red, two green and one blue filter (RGGB): 50 percent of the pixels are green, because the eye is most sensitive to green and most of the brightness detail is in green, and 25 percent each red and blue. Each pixel records only one colour, and the two missing colour values at each pixel are estimated from neighbouring pixels by interpolation, called demosaicing. Professional video cameras instead use a beam-splitting prism with three separate sensors for R, G and B (3-chip cameras), and some sensors stack three photodiodes at different depths, using the fact that blue light is absorbed near the surface and red light deeper. The final colour image is stored as three numbers (R, G, B) per pixel.
Colour and the CIE system: Human colour vision uses three kinds of cone cells sensitive to long (red), medium (green) and short (blue) wavelengths, so any colour can be described by three numbers. In 1931 the CIE (International Commission on Illumination) defined three imaginary primaries X, Y and Z and colour matching functions \(\bar{x}(\lambda), \bar{y}(\lambda), \bar{z}(\lambda)\). The tristimulus values X, Y and Z of a light are found by multiplying its spectrum by these functions and adding over all wavelengths. Y was chosen to equal the luminance (brightness) of the colour. All three values are always positive, unlike matches made with real red, green and blue lamps, which need negative amounts for some colours.
Chromaticity coordinates and diagram: To separate colour quality from brightness, the tristimulus values are normalised: \(x = X/(X+Y+Z)\), \(y = Y/(X+Y+Z)\), \(z = Z/(X+Y+Z)\), so \(x + y + z = 1\) and only x and y need be given. The colour is then specified by (x, y) plus the luminance Y. Plotting y against x for all colours gives the CIE 1931 chromaticity diagram, a horseshoe-shaped region. Its curved boundary, the spectral locus, holds the pure spectral colours from about 380 nm (violet, bottom left) through 520 nm (green, top) to 700 nm (red, right). The straight line joining the two ends is the line of purples (magentas, which are not spectral colours). White lies near the centre: the equal-energy white point E is at (1/3, 1/3), and daylight D65 at about (0.3127, 0.3290). Colours become more saturated towards the boundary. Any mixture of two lights lies on the straight line joining their points, and the colours that three primaries can produce lie inside the triangle (gamut) formed by their points; for example the sRGB display primaries are at about R (0.64, 0.33), G (0.30, 0.60) and B (0.15, 0.06). No triangle of real primaries can cover the whole horseshoe, which is why no screen shows every colour. The dominant wavelength of a colour is found by drawing a line from the white point through the colour to the spectral locus, and its purity is its distance from white relative to the distance of the locus.
Colour measurement: A colorimeter measures X, Y, Z directly using three detectors fitted with filters shaped like the colour matching functions; it is quick and cheap. A spectrophotometer (or spectroradiometer for light sources) measures the full spectrum wavelength by wavelength and calculates X, Y, Z from it; it is more accurate and can detect metamerism. For surfaces, a standard illuminant (for example D65) and standard observer are used. To express colour differences in a way that matches human perception, the uniform colour space CIELAB (L* lightness, a* red-green, b* yellow-blue) is used, and the difference between two colours is \(\Delta E^*_{ab} = \sqrt{(\Delta L^*)^2 + (\Delta a^*)^2 + (\Delta b^*)^2}\); a value around 1 to 2 is just noticeable, while larger values show a visible mismatch. Colour measurement is used in quality control of paints, textiles, printing, food, displays and LEDs.
Colour matching: In the classic colour matching experiment an observer looks at a split (bipartite) field: one half shows the test colour and the other half a mixture of three primary lights (red, green and blue) whose intensities the observer adjusts until both halves look the same. The amounts of the three primaries needed are the tristimulus values of the test colour. Grassmann's laws state that colour matching is additive and linear: if A matches B and C matches D, then A + C matches B + D, and scaling both sides keeps the match. Some spectral colours (such as bright blue-green) cannot be matched by adding the primaries; instead some red must be added to the test side, which counts as a negative amount, and this is why CIE introduced the all-positive XYZ system. Two lights with different spectra that look identical are called metamers; metamerism makes colour matching possible with only three primaries, but it also means two samples that match under one light may not match under another. In practice, cameras are colour-calibrated against charts of known colours, and screens are calibrated so that their RGB values give the intended chromaticity.
Mind map
Optical Imaging
- Imaging sensors
- Pixel array, photoelectric effect
- Charge ∝ intensity × time
- ADC gives grey levels
- CCD
- MOS capacitor wells
- Three-phase clocking
- Serial register, one output
- Full-frame, frame-transfer, interline
- CMOS (APS)
- Photodiode + 3T/4T per pixel
- Row-column addressing
- On-chip ADC, low power
- Rolling or global shutter
- Image formation
- Monochrome: grey levels 0 to 255
- Colour: Bayer RGGB + demosaicing
- 3-chip prism cameras
- Colorimetry
- CIE XYZ, Y = luminance
- x = X/(X+Y+Z), y = Y/(X+Y+Z)
- Horseshoe diagram, white point
- Gamut triangle
- Measurement and matching
- Colorimeter, spectrophotometer
- CIELAB ΔE
- Grassmann's laws, metamerism
Readout in a CCD versus a CMOS sensor
- Both sensors: the lens forms an image; photons absorbed in each pixel free electrons, which are stored during the exposure time.
- CCD: at the end of exposure, the gate voltages are clocked so that every charge packet moves one pixel down its column.
- CCD: the bottom row is dumped into the horizontal serial register, which shifts packets one at a time to a single output amplifier.
- CCD: the amplifier converts each packet into a voltage, and an external ADC digitises it; the process repeats row by row.
- CMOS: each pixel converts its own charge into a voltage with its source-follower transistor.
- CMOS: the row decoder switches on one row-select line; all pixels of that row place their voltages on the column lines at once.
- CMOS: column circuits sample (with correlated double sampling to remove reset noise) and column ADCs digitise the row in parallel.
- Result: the CCD moves charge across the chip to one output (uniform, low noise, slow); the CMOS sensor converts charge in place and reads in parallel (fast, low power, integrated).
Important definitions
| Term | Meaning |
|---|---|
| Pixel | The smallest light-sensing element of an image sensor and the smallest picture element of a digital image. |
| CCD | Charge-coupled device: an array of MOS capacitors in which photo-charge is stored in potential wells and shifted from pixel to pixel to a common output. |
| CMOS image sensor (APS) | An image sensor in which each pixel has a photodiode and its own transistors for conversion and addressing, read row by row. |
| Charge transfer efficiency | Fraction of a charge packet successfully moved from one CCD gate to the next in one transfer. |
| Fill factor (sensor) | Fraction of a pixel area that is light-sensitive. |
| Quantum efficiency | Number of electrons collected per incident photon at a given wavelength. |
| Bayer filter | A colour filter array with a repeating 2 × 2 RGGB pattern placed over the sensor pixels. |
| Demosaicing | Interpolation that estimates the two missing colour values at each pixel of a Bayer sensor. |
| Tristimulus values | The amounts X, Y, Z of the three CIE primaries needed to match a colour. |
| Chromaticity coordinates | Normalised values x = X/(X+Y+Z), y = Y/(X+Y+Z), z = 1 − x − y that describe colour quality independent of brightness. |
| Chromaticity diagram | Plot of y against x showing all visible colours inside the horseshoe-shaped spectral locus. |
| Gamut | The range of colours a device can produce; for three primaries, the triangle joining their chromaticity points. |
| Metamers | Lights or surfaces with different spectra that appear the same colour under given conditions. |
| Dominant wavelength | The spectral wavelength at which a line from the white point through the colour meets the spectral locus. |
Formulas
| Formula | Meaning |
|---|---|
| \( X = \sum P(\lambda)\bar{x}(\lambda)\Delta\lambda,\ \ Y = \sum P(\lambda)\bar{y}(\lambda)\Delta\lambda,\ \ Z = \sum P(\lambda)\bar{z}(\lambda)\Delta\lambda \) | Tristimulus values from the spectrum P(λ) and the colour matching functions. |
| \( x = \dfrac{X}{X+Y+Z},\ \ y = \dfrac{Y}{X+Y+Z},\ \ z = \dfrac{Z}{X+Y+Z} \) | Chromaticity coordinates. |
| \( x + y + z = 1 \) | So only (x, y) are needed. |
| \( X = \dfrac{x}{y}Y,\ \ Z = \dfrac{1 - x - y}{y}Y \) | Recovering X and Z from (x, y, Y). |
| \( (X, Y, Z)_{mix} = (X_1 + X_2,\ Y_1 + Y_2,\ Z_1 + Z_2) \) | Additive mixing of two lights (Grassmann's law). |
| \( \Delta E^*_{ab} = \sqrt{(\Delta L^*)^2 + (\Delta a^*)^2 + (\Delta b^*)^2} \) | Colour difference in CIELAB. |
| \( N_{levels} = 2^n \) | Number of grey levels for an n-bit ADC. |
| \( \text{Image size (bytes)} = M \times N \times \dfrac{n}{8} \times c \) | Uncompressed size for M × N pixels, n bits per channel, c channels (1 for grey, 3 for RGB). |
| \( \lambda_c = \dfrac{hc}{E_g} \approx 1.1\ \mu\text{m} \) | Longest wavelength detected by a silicon sensor. |
Key points to remember
- Both CCD and CMOS sensors use the photoelectric effect in silicon; they differ in readout.
- CCD: charge is moved through the chip to one output amplifier; CMOS: charge is converted to voltage inside each pixel.
- CCD needs multiple clock voltages and more power; CMOS works at low voltage with on-chip ADC and processing.
- The charge collected is proportional to intensity × exposure time, until the pixel saturates (full-well capacity).
- 8-bit images have 256 grey levels; colour images have three channels.
- Bayer pattern: 50 percent green, 25 percent red, 25 percent blue pixels.
- In CIE XYZ, Y is the luminance; X, Y, Z are always positive.
- x + y + z = 1, so a colour is given by (x, y) and Y.
- Equal-energy white is at (1/3, 1/3); D65 daylight is at (0.3127, 0.3290).
- Spectral colours lie on the horseshoe boundary; purples lie on the straight line of purples.
- Mixtures of two colours lie on the straight line joining them; three primaries give a triangular gamut.
- Metamers look alike but have different spectra; colour matching depends on this.
Solved examples
Example 1. A colour has tristimulus values X = 30, Y = 40, Z = 30. Find its chromaticity coordinates.
\(X + Y + Z = 100\). \(x = 30/100 = 0.30\), \(y = 40/100 = 0.40\), \(z = 30/100 = 0.30\). Check: \(x + y + z = 1\). The point (0.30, 0.40) lies a little above white towards green, so the colour is a pale greenish white with luminance Y = 40.
Example 2. D65 white has x = 0.3127, y = 0.3290. If its luminance is Y = 100, find X and Z.
\(X = (x/y)Y = (0.3127/0.3290) \times 100 = 95.05\). \(z = 1 - 0.3127 - 0.3290 = 0.3583\), \(Z = (z/y)Y = (0.3583/0.3290)\times100 = 108.9\). So D65 white is (X, Y, Z) = (95.05, 100, 108.9).
Example 3. Light A has (X, Y, Z) = (20, 30, 10) and light B has (10, 10, 40). Find the chromaticity of the mixture.
By Grassmann's law add the tristimulus values: X = 30, Y = 40, Z = 50, total 120. \(x = 30/120 = 0.25\), \(y = 40/120 = 0.333\). The mixture point lies on the straight line joining A (0.333, 0.5) and B (0.167, 0.167) on the chromaticity diagram.
Example 4. A full-HD sensor has 1920 × 1080 pixels and is 6.4 mm wide. Find the number of pixels, the pixel pitch, and the number of green, red and blue pixels in its Bayer filter.
Pixels \(= 1920 \times 1080 = 2{,}073{,}600 \approx 2.07\) megapixels. Pixel pitch \(= 6.4\ \text{mm}/1920 = 3.33\) μm. Green pixels \(= 50\% = 1{,}036{,}800\); red \(= 25\% = 518{,}400\); blue \(= 518{,}400\).
Example 5. Find the uncompressed size of a 1920 × 1080 image stored (a) as an 8-bit greyscale image and (b) as a 24-bit RGB image.
(a) \(1920 \times 1080 \times 1\) byte \(= 2{,}073{,}600\) bytes \(\approx 2.07\) MB. (b) \(1920 \times 1080 \times 3\) bytes \(= 6{,}220{,}800\) bytes \(\approx 6.22\) MB. An 8-bit channel gives \(2^8 = 256\) levels, so each pixel in (b) can show \(256^3 \approx 16.7\) million colours.
Example 6. A pixel has a full-well capacity of 20000 electrons and a quantum efficiency of 50 percent. How many photons saturate it?
Each photon gives on average 0.5 electron, so photons needed \(= 20000/0.5 = 40000\). More light than this during the exposure produces no further signal (the pixel is clipped to white).
Example 7. Two paint samples have CIELAB values (L*, a*, b*) = (50, 20, 10) and (52, 17, 14). Find ΔE and comment.
\(\Delta L^* = 2,\ \Delta a^* = -3,\ \Delta b^* = 4\). \(\Delta E^*_{ab} = \sqrt{4 + 9 + 16} = \sqrt{29} = 5.39\). This is well above the just-noticeable difference of about 1 to 2, so the two samples do not match; the batch would be rejected in quality control.
Important questions and answers
Q1. Explain the construction and working of a CCD image sensor. (5 marks)
Construction: p-type silicon substrate, thin SiO2 layer, rows of transparent polysilicon gates (three per pixel), channel stops between columns, a horizontal serial register at the bottom and an output amplifier at one corner. Working: (i) Integration: a positive voltage on one gate of each pixel forms a potential well; photo-electrons generated by incident light collect in it, the charge being proportional to light intensity and exposure time. (ii) Transfer: the three gate voltages are switched in sequence so that each well moves under the next gate; after three clock steps every packet has moved one pixel down the column; the bottom row enters the serial register. (iii) Readout: the serial register shifts packets one by one into the output node, where the charge is converted into a voltage, amplified and digitised. Diagram: cross-section with gates φ1, φ2, φ3 over oxide and the potential wells shown below them, plus a top view of the array, serial register and amplifier.
Q2. Explain the construction and working of a CMOS image sensor. (5 marks)
Each pixel contains a photodiode and transistors: reset (RST), transfer gate (TX, in 4T pixels), source follower (SF) and row select (SEL). The pixel array is surrounded by a row decoder, column amplifiers with correlated double sampling, column ADCs and timing control. Working: RST empties the photodiode; during exposure, light produces electrons that collect in the photodiode; TX transfers them to the floating diffusion, whose voltage drops in proportion to the charge; SF buffers this voltage; when the row is selected, SEL connects SF to the column line; the column circuit samples the reset level and the signal level, subtracts them and digitises the result. Rows are read one after another. Diagram: a 4T pixel schematic and a block diagram of the array with row decoder and column ADCs.
Q3. Compare CCD and CMOS image sensors. (5 marks)
Readout: CCD shifts charge to a single output; CMOS converts charge in each pixel and reads by row-column addressing. Power: CCD high (several clock voltages); CMOS low (single low supply). Speed: CCD slower (serial); CMOS faster (parallel columns). Integration: CCD needs external ADC and processing chips; CMOS has them on the same chip (camera on a chip). Noise and uniformity: CCD traditionally lower noise and very uniform; CMOS has some fixed-pattern noise but modern sensors are comparable. Fill factor: CCD high; CMOS lower because of in-pixel transistors (improved with microlenses and backside illumination). Random access/windowing: not possible in CCD; possible in CMOS. Shutter: interline CCD has a global electronic shutter; most CMOS use a rolling shutter. Cost: CCD needs a special process; CMOS uses standard CMOS fabrication and is cheaper. Uses: CCD in astronomy and scientific cameras; CMOS in phones, DSLRs, webcams, cars, drones.
Q4. How are monochrome and colour images formed by an image sensor? (5 marks)
Monochrome: the lens forms an image on the pixel array; each pixel collects charge proportional to the light reaching it; the readout circuit converts this to a voltage; the ADC converts the voltage into a grey level (0 to 255 for 8 bits); the set of grey levels arranged in rows and columns forms the digital image f(x, y). Colour: since silicon cannot sense colour, a colour filter array is placed over the pixels. With the Bayer RGGB pattern each pixel records only red, green or blue; demosaicing interpolates the missing two values from neighbours to give R, G and B at every pixel. Alternatives: a three-sensor camera with a dichroic prism that splits light into R, G and B beams, each falling on its own sensor (better colour, bulky), or stacked photodiodes that use the depth of absorption. White balance and colour correction then adjust the RGB values to match the scene.
Q5. Explain the CIE chromaticity diagram. (5 marks)
The CIE 1931 system describes any colour by tristimulus values X, Y, Z, with Y equal to luminance. Normalising gives chromaticity coordinates \(x = X/(X+Y+Z)\) and \(y = Y/(X+Y+Z)\). The plot of y versus x for all visible colours is a horseshoe. Features: the curved spectral locus carries pure spectral colours from 380 nm to 700 nm labelled in nm; the straight line of purples joins its ends; the white point lies near (1/3, 1/3); saturation increases from the white point towards the boundary; all real colours lie inside; a mixture of two colours lies on the line joining them; the gamut of an RGB display is the triangle joining its primaries, which covers only part of the horseshoe. Uses: specifying colours of LEDs, displays and signal lights, comparing gamuts, finding dominant wavelength and purity.
Q6. What are chromaticity coordinates? Why is z not needed? (3 marks)
Chromaticity coordinates are the fractions of each tristimulus value in their sum: \(x = X/(X+Y+Z)\), \(y = Y/(X+Y+Z)\), \(z = Z/(X+Y+Z)\). They describe hue and saturation independent of brightness. Because \(x + y + z = 1\), z can always be found as \(1 - x - y\), so two coordinates are enough to locate a colour on a plane diagram. Brightness is given separately by Y.
Q7. Explain the colour matching experiment and Grassmann's laws. (5 marks)
An observer views a circular split field. One half is lit by the test colour; the other by a mixture of red, green and blue primary lights with adjustable intensities. The observer adjusts the three intensities until the halves are indistinguishable. The settings are the tristimulus values of the test colour. Repeating this for each spectral wavelength gives the colour matching functions. For some wavelengths a match is impossible unless one primary (usually red) is added to the test half, which is recorded as a negative value; CIE therefore transformed the data to imaginary primaries X, Y, Z with all-positive matching functions. Grassmann's laws: (i) three independent variables are necessary and sufficient to specify a colour match; (ii) additivity: if A matches B and C matches D then A + C matches B + D; (iii) proportionality: if A matches B then kA matches kB. These laws make colorimetry linear.
Q8. How is colour measured? (3 marks)
A tristimulus colorimeter uses three photodetectors behind filters that mimic \(\bar{x}, \bar{y}, \bar{z}\) to give X, Y, Z directly. A spectrophotometer measures the reflectance or transmittance at each wavelength and computes X, Y, Z for a chosen illuminant and observer. From these, chromaticity (x, y) and CIELAB values are calculated, and the colour difference ΔE between a sample and a standard is used to accept or reject products.
Q9. What is metamerism? Give its significance. (3 marks)
Metamerism is the phenomenon in which two colour stimuli with different spectral power distributions appear identical to an observer under particular conditions. It occurs because the eye reduces a full spectrum to only three cone responses. Significance: it is the basis of colour reproduction (a screen shows yellow using red and green light instead of spectral yellow), but it also causes problems such as two fabrics matching in the shop light but not in daylight, so colour matching in industry is checked under several standard illuminants.
Exam tips
- For CCD draw the three-phase gates with wells and show one packet moving; examiners look for the clocking idea.
- For CMOS draw the 3T or 4T pixel and name each transistor's job.
- Learn at least six points of CCD versus CMOS in a table.
- Draw the horseshoe neatly: label 380, 480, 520, 560, 600 and 700 nm, the line of purples, white point and an RGB triangle.
- In chromaticity numericals always check that x + y + z = 1.
- For mixing, add X, Y, Z first, then find x and y; never average x and y directly unless luminances are weighted.
- Mention Grassmann's laws and metamerism in any colour matching answer.
FAQs
Why does a Bayer filter have twice as many green pixels?
The human eye is most sensitive to green light, and green carries most of the brightness detail, so sampling green more often gives sharper looking images.
Can a camera sensor see infrared?
Yes. Silicon responds up to about 1.1 μm, so cameras include an IR-cut filter to keep infrared from spoiling colours. Night-vision cameras remove it.
Why can no display show all the colours of the chromaticity diagram?
A display mixes three real primaries, and mixtures lie only inside their triangle. Since the horseshoe boundary is curved, no triangle with real corners can enclose it completely.
Is the Y in XYZ the same as yellow?
No. Y is the luminance (brightness) of the colour. X, Y and Z are imaginary primaries, not real red, yellow or blue lights.
These notes follow the University of Mumbai first-year engineering syllabus (NEP 2020, Physics for Emerging Fields, Semester II) and were prepared with AI help and checked.