Colour Science & Perception
what the eye settles, and what it cannot
Theme

The thread: The instrument is the reader — page 2

This is the one subject where the page is displayed on the apparatus under discussion, and the reader's own eye is the measuring device. Several figures here are experiments rather than illustrations.
The best a camera profile can do on the surfaces it was fitted to. Per-surface ΔE00 after the best 3 × 3 from raw to XYZ, fitted on 12 surfaces at chroma 0.7 and tested on those same 12 surfaces — mean 1.45, worst 2.58. This is the most favourable measurement it is possible to make of a camera and it is the one usually published. What a camera does

No matrix is right everywhere

The best three-by-three this sensor admits, fitted and tested on the same twenty-four surfaces, leaves a worst case of ΔE00 2.85. A control sensor built to satisfy Luther's condition reaches ten to the minus seven under the identical computation, which is what makes the first number a measurement.

 A grey edge, reconstructed from a Bayer row, arrives coloured. Above: an achromatic step through 24 sensor sites, with green sampled on the even ones and red on the odd. Interpolating each channel separately reconstructs them from data taken on either side of the edge, so their ratio moves. Below: the resulting chroma, peaking at 144 per cent of the local mean, and 112 per cent once colour differences are interpolated instead. What a camera does

A grey edge arrives coloured

A sensor site measures one channel and the other two are interpolated from neighbours that sat somewhere else. Across a black-and-white step that reconstruction gives an achromatic scene a chroma of 144 per cent of its own local mean, and nothing in the scene or the sensor was coloured.

 Recovering the lamp from a highlight rather than from an assumption. A green scene under illuminant A. The faint curve is the true lamp; the solid one is what the interface component of the glossy surfaces recovers, which is 0.30° from it. Grey-world on the same scene is 42.1° and max-RGB 17.5°, because both are assumptions about the surfaces and a Fresnel reflection is not. What a camera does

The highlight is the white balance

Every estimator a camera uses is an assumption about the scene wearing the costume of a measurement. The specular component of a glossy surface is not — it carries the illuminant to within a third of a degree on a scene where averaging is fifty times worse, and it does so without knowing what colour the paint underneath is.

 Correcting colour costs noise, and the two cannot both be least. Sweeping the colour matrix from its own diagonal — a white balance with no cross terms — to the full least-squares fit. Mean ΔE00 falls from 18.61 to 1.29; photon noise rises by a factor of 1.03. At 10000 photons per pixel. What a camera does

Correcting colour costs noise

The matrix that turns a sensor's raw into XYZ has large off-diagonal terms of both signs, because that is what correcting a sensor which fails Luther's condition requires. Differences of large numbers are where noise grows, and the full correction amplifies photon noise by 1.66 times.

 A clipped channel turns the hue of what is left. CIELAB hue shift against exposure for one saturated stimulus, measured against the same stimulus rendered without clipping. Nothing moves until the first channel reaches the ceiling at 0.25 stops; after that the recorded hue rotates by as much as 67 degrees, with nothing in the scene having changed colour. What a camera does

A blown highlight turns

An exposure that clips nothing changes no hue at all. Once one channel reaches the ceiling the recorded hue rotates by as much as sixty-seven degrees, with nothing in the scene having changed colour — and every response a converter can make to that is an invention.

 A camera's spectral sensitivities, with the filter removed. Silicon quantum efficiency times the colour-filter dye times nothing else, per channel, on a grid running to 1100 nm rather than to 780. With the filter removed, 68 per cent of the area under the three curves lies beyond the visible band, and all three curves are the same curve out there. What a camera does

The grid outside every figure

Every figure here is computed on 380 to 780 nanometres, which is exactly right for an eye and insufficient for a sensor. This field met the first subject the grid cannot hold, and the decision was not to widen it — because widening it honestly is impossible.

 Everything between the photons and the picture, and what each stage decides. The 8 stages of a camera pipeline. Only the second is physics; every one after it is a decision somebody made, and the reason two cameras pointed at the same scene disagree is that they made different ones. What a camera does

A photograph is not a measurement

A photograph is a measurement made by an instrument whose kernel nobody published, under an illuminant nobody recorded, corrected by a matrix fitted to somebody else's surfaces, with two thirds of every pixel invented. It supports relative claims well and absolute ones badly, and it is used for the second.

 A halftone tint, its four regions, and what they average to. At 40% and 30% coverage the sheet is a mosaic of 4 fully-inked regions, not a mixture of anything. Their areas are the product of the coverages — Demichel's rule, which holds because the screens are rotated to make it hold — and the patch's reflectance is the area-weighted average of theirs, raised to the Yule–Nielsen exponent n = 1.8. Averaging the regions' CIELAB coordinates instead, which is what mixing means to almost everybody, lands ΔE00 = 0.76 away. What it takes to deliver it

A halftone is not a mixture

Forty per cent cyan and thirty per cent magenta are not stirred together anywhere. They are laid down as dots, and the sheet is a mosaic of four fully-inked regions in proportions the two coverages fix — which is why the colour is an area average of four spectra rather than a blend of two, and why it lands nowhere near where mixing would put it.

 Two mechanisms, one ramp, and the patch that separates them. A cyan ramp printed by a press with 10 per cent mechanical gain and a Yule–Nielsen exponent of 2.4. Fitting a mechanical gain at n = 1.0 gives 23 per cent and fitting one at n = 2.4 gives 10 per cent; both reproduce the ramp to under 0.66 of a lightness unit, and the curves lie on top of one another. The two-ink overprint below was in neither fit, and there they are ΔE00 = 3.20 apart. What it takes to deliver it

A dot is larger than it was asked to be

Two quite different things make a printed midtone darker than its coverage says — ink spreading under pressure, and light scattering sideways inside the paper. Both bend the tone curve the same way and neither touches its ends, so the measurement every press is characterised by cannot say which happened. A patch that nobody measures can.

 How wrong a profile is between its entries. A device profile is a lattice of measured patches and an interpolation. At every node of every grid drawn here the error is exactly zero, which is why a profile checked at its own patches always looks perfect. Sampled between the nodes, a 3-step grid is worst by ΔE00 = 2.93 and a 17-step grid by 0.30. Both are honest tables of the same press. What it takes to deliver it

A profile is a table

An ICC profile does not contain a model of the device it describes. It contains a lattice of patches that were printed and measured, and everything between them is interpolation — which is exactly zero at every patch, so a profile checked against its own target reports perfection for ever.

 What separates the separations, and under which light. 11 separations of a single colour, differing only in how much of it is carried by black rather than by the three chromatic inks. Under D50 they agree to ΔE00 = 0.00 — the solver was asked for that and delivered it. Under illuminant A they spread to 6.75. They are metamers of one another, and the whole family is invisible to any instrument reading a single illuminant. What it takes to deliver it

The lamp in the shop decides

Two packages printed to the same specification, verified under the same standard illuminant and passed, can differ by nearly seven colour differences on a shelf under a tungsten lamp. Nothing went wrong at either press. The specification named one light, the separations that satisfy it are metamers of one another, and the shop chose the second light.

 The tone scale of one display, in four rooms. Code value along the bottom, the lightness it comes out at up the side. The curves are the same display: what changes is the light falling on the screen, which adds a fixed luminance to every pixel. The signal value that means black arrives at L* = 11.7 at 600 lux, against 0.9 in the dark — the lightness of a dark grey card, from a pixel that is switched off. What it takes to deliver it

The screen is not the room

A thousand-to-one display in an ordinary office is a hundred-and-thirty-seven-to-one display, and the signal value that means black comes out at the lightness of a dark grey card. Nothing is wrong with the panel. Light falls on the screen and comes back off it, and the two per cent it returns is added to every pixel — most of all to the ones with nothing else in them.

 An edge at 4:2:2, and the colour it arrives as. Above, the row as it was sent and as it arrives after the two chroma planes are averaged 2 samples at a time. Below, the colour difference at each sample. The worst is ΔE00 = 25.88, at a luma step of 0.058 across the edge — an edge of nearly equal luminance. Nothing is wrong with the codec: it discards the differences the eye resolves worst, and this edge is made of nothing else. What it takes to deliver it

Colour thrown away on purpose

Every video format in use discards three quarters of its colour information and keeps all of its luminance, because the eye resolves fine colour detail badly. On an edge that carries luminance the loss is exactly zero. On an edge of nearly equal luminance it is a colour difference of forty-three, and the two edges are the same edge to the codec.

 D65, and a source built to have its chromaticity and nothing else. The smooth curve is the CIE's daylight reconstruction at 6504 K. The other is five Gaussian emission bands whose weights were solved so that the two agree in chromaticity to 8.5e-10 — closer than any instrument could tell them apart when looking at the lamps. Three numbers were matched and seventy-eight were not, and everything either source falls on will report the difference. What light is

There is no D65 lamp

D65 is standardised by three numbers, and three numbers do not pin a spectrum. A source built to hit them exactly — matched to a part in a billion, with a third of the visible band nearly empty — separates pairs that D65 says are identical by up to eight and a half units of colour difference.

 Contrast sensitivity, three channels, normalised to each channel's peak. Spatial frequency in cycles per degree against relative sensitivity. The luminance channel runs to 50 cycles per degree, red–green to 12 and blue–yellow to 8, each by the same criterion of 5 per cent of that channel's own peak. Only the luminance channel dips at low frequency; the two chromatic ones are low-pass, so a chromatic edge of any size at all is seen at its full contrast. What the eye does

How fine a colour edge can be

The eye resolves a lightness pattern to about fifty cycles per degree and a red–green one to twelve. Every colour difference here is quoted as though a patch had no size, and the same difference is plainly visible at one scale and gone at another.

 Photons caught by one cone in one integration time, under D65. Each curve is one cone class, counting isomerisations during a 0.1 s integration through a pupil that closes as the light rises. At 1000 cd/m² a long-wavelength cone catches about 11522 and at 0.001 it catches 0.12 — which is where the square root of the count stops being a small correction and starts being the signal. The rate works out at 30 isomerisations per second per troland, inside the published band of 5–50. What the eye does

Colour goes first in the dark

A cone reports a count, and a count carries the square root of itself as noise. Counting the photons says where colour vision stops — a chromatic difference runs out four hundred times sooner than a lightness difference of the same size — and says just as clearly that in daylight the eye is nowhere near that limit.

 What a coarse wavelength grid costs, by source. Colour error against grid size, for four sources through one reflectance. Daylight survives every grid tested: 0.67 ΔE00 even at 40 nm. A source with lines in it does not — the narrowband source reaches 16.2. The grid is not a property of the arithmetic; it is a claim about what the light has in it. Matching and measuring

Five nanometres is a choice

Every integral here is taken in five-nanometre steps, and the interval has never had to be defended. Coarsening it to twenty costs daylight two hundredths of a colour difference and a fluorescent tube six and a half — and which way of coarsening is used decides a further factor of six.

 A chromatic line screen at 10.0 cycles per degree. The upper strip is the pattern as delivered and the lower one is the same pattern after each opponent channel has been low-passed at its own cutoff. Against a flat field of the same mean, the delivered pattern differs by up to ΔE00 14.22 and the filtered one by 7.77 — a ratio of 1.8. Every number quoted elsewhere for a difference of this kind is the first one. Difference and uniformity

A difference has no size

A colour difference formula answers a question about two large patches seen side by side. Applied to a pattern, it reports fourteen units where the eye is left with less than one — and the ratio depends on nothing but how finely the difference is spread.

 Brightness against chroma, at exactly constant luminance. Seven stimuli of identical luminance and rising chroma at hue 25. The model's brightness moves by 2.2 per cent across the whole sweep, and at some hues it moves the other way. The Helmholtz–Kohlrausch effect — measured repeatedly, by several methods — is that a saturated colour looks as bright as a neutral of 1.3 to 2 times its luminance, shown as the band. The gap is the model's, and nothing here closes it: the term that would is not in CIECAM16 and is not invented for the occasion. What the brain does

Brightness is not luminance

Photometry is additive because the CIE defined it that way. Brightness is not, and a saturated colour looks as bright as a neutral of one and a half times its luminance — an effect this site's appearance model moves by two per cent, in a direction that depends on the hue.

 A 8-bit ramp from 0.0008 to 0.006 of white, 12° wide. The top strip is the ramp as delivered: 16 distinct levels, each a step of one code value. Below it is the quantisation error as a Weber contrast against the local luminance, filtered by the luminance sensitivity function. The largest response is 8.51 per cent contrast against a threshold of 0.3, which is 28.4 times over — and it occurs at 0.09 per cent of white, at the dark end, because a code step is a Weber contrast and the same step is a larger fraction of less light. Where the model breaks

Banding is not a bit depth

Eight bits bands at thirty times threshold in the shadows and at under three at mid grey, on the same ramp with the same encoding. What decides is where in the tone scale the gradient sits, how sharp each step's edge is, and how far away the reader is — and a bit count contains none of the three.

 How far a judgement is from the settled one, second by second. The light changed from one white to another at t = 0 and nothing else moved. The model has one degree of adaptation and no clock, so the distance plotted is what a clock adds: 3.6 CAM16-UCS units half a second in, still 1.3 after a minute, and 0.11 after five. Every appearance number on this site is the value at the right-hand end. Where the model breaks

The model has no clock

An appearance model takes a stimulus and a situation and returns what it looks like. It does not take a time, and adaptation is not instantaneous — half a second after the light changes a judgement is three and a half CAM16-UCS units from the settled one, and a minute later it is still 1.3.

 A screen at 45°, 8 c/°, and where its energy sits. Left, the pattern. Right, its power in the frequency plane with the zero frequency at the centre and the edges at the sampling limit of 23 cycles per degree, on a logarithmic scale over five decades. The closed curves are the visual system's own sensitivity at 5, 25, 60 per cent of its peak; they are not circles, because sensitivity is lower on the diagonals than on the cardinal axes by a factor of 2.0 at high frequency. Energy inside a curve is seen; energy outside it is not, whatever its size. What the eye does

A pattern has a direction

Every spatial claim here is a claim about a frequency, and a frequency has no direction in it. Turning a printed screen forty-five degrees makes it exactly twice as quiet with nothing else changed — and the same rotation does nothing at all to a chromatic one.

 Cone density, out from the centre of gaze. Quoted landmarks with logarithmic interpolation between them: 199,000 cones per square millimetre at the fovea, 9,500 at ten degrees — a factor of 21. The shaded bands are the two standard observers' fields. The 2° observer averages over a region whose density falls by 3.3× between its centre and its edge and the 10° observer by 12.4×, which is what "a 10° field" contains and is why the two sets of matching functions are different shapes rather than the same shape scaled. What the eye does

Colour stops at the edge of sight

Cone density falls twenty-one-fold between the centre of gaze and ten degrees out, and the three channels give out at three different rates — so a colour difference in the periphery does not merely shrink, it turns. At the exact point of fixation there are no short-wavelength cones at all.

 Temporal sensitivity, and where each channel gives out. Modulation frequency in hertz against relative sensitivity. The luminance channel is band-pass, peaking at 8 hertz and running out at 60; an isoluminant modulation is low-pass and runs out at 15, which is 4.0 times sooner. Both cutoffs are at the same criterion of 5 per cent of that channel's own peak, so the ratio between them is a ratio between two measurements rather than between two conventions. What the eye does

The eye has a shutter

An isoluminant flicker fuses at fifteen hertz and a luminance one at sixty, so a light whose colour changes forty times a second is a steady light of a colour it never emits. And the frequency at which flicker stops being visible is not a property of the eye — it moves twelve and a half hertz for every decade of light.

258 essays on this thread, page 2 of 11 · all threads · all essays