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

The thread: Three numbers — page 9

An infinite-dimensional spectrum is projected onto three cone responses, and everything colour science can do — and every way it fails — follows from that single collapse.
Two ways to ask how much rod signal an older eye has. The rod signal's catch of each lamp divided by the S cones' own catch, against the observer's age, for five lamps — drawn twice. The upper curves take the rod signal at a fixed absolute size, and every one of them roughly doubles from twenty to seventy-five: an older lens cuts the blue before the S cones see it and the rods, peaking further into the green, lose much less. The lower curves take the rod signal at a tenth of each cone's own peak absorptance, which is the model's own definition, and they barely move at all. Nothing about the retina differs between the two; only the normalisation does. What the eye does

A rod signal has no natural size

An older lens absorbs where the S cones are sensitive, so it should make the uncertain rod-to-S-cone weight cheaper. Measured, the rod signal's catch of a phosphor LED as a share of the S cones' own catch more than doubles from twenty to seventy-five — and the share the model actually uses falls by a fifth. The two differ by the S cone's peak absorptance, which the lens takes 60 per cent of, and which entered the model as a normalisation rather than as a claim.

 What survives holding the lightness still. For each of five groups of rooms sorted by how much light the wall returns, the rank correlation of the finish's share with three properties of the paint, taken with the wall's luminance factor held. The wall's chroma runs from -0.90 among the darkest walls to 0.81 among the palest, crossing zero in the middle — the reversal. The band's width, which was the predicted mechanism, never leaves the range -0.15 to 0.22, and the census already contains four families whose bands are all the same width. What a scene does

The finish adds the room's own colour

Among dark walls a more saturated paint loses a smaller share of its colour to a gloss finish, and among pale walls a larger one. The mechanism proposed for that reversal was spectral concentration — a narrow tall band — and the census that found it already contained four families of paints whose bands are all the same width. Holding lightness still, the band's width orders the losses at a rank correlation of 0.01. What does order them is that the light a finish adds has already bounced off the walls.

 How each unit balances lightness against chroma, as a display brightens. Twelve pairs built to differ in lightness alone and twelve built to differ in chroma alone, each at exactly one ΔE₀₀, read in both units at nine display levels. Up is the median lightness pair's reading divided by the median chroma pair's, so a falling curve means chroma differences becoming relatively more visible. Both fall. ΔEITP's median runs from 0.91 at a 1.5-candela white to 0.81 at 10,000, a fall of 11 per cent; CAM16-UCS's from 1.13 to 0.87, a fall of 23 per cent. Neither turns: both units say chroma differences gain on lightness differences as a display brightens, and CAM16-UCS says it twice as strongly. Difference and uniformity

The units part by hue, not by light level

Two colour differences with a light level in them were compared on a display running from 5 to 5,000 cd/m², and a prediction was drawn from how each divided a difference between lightness and chroma. Built into pairs that differ in lightness alone and in chroma alone, both units say the same thing about light level: as a display brightens, chroma differences gain on lightness differences — ΔEITP by a tenth, CAM16-UCS by a quarter. Where they part is hue. The appearance model moves every colour's balance by the same factor; ΔEITP moves the violets, reds and cyan-blues the other way, and at nine of twelve base colours the two units disagree about the direction.

 What a lit room takes, and where it takes it. On a display whose white is 1,000 cd/m², how much of the ΔEITP a one-unit lightness step is given survives a veiling luminance reflected off the screen, against where on the lightness scale the step sits. At L 2 — a deep shadow — one candela of reflected light removes 22 per cent of the difference and three candelas remove 46. At L 90 the same veils remove nothing measurable. The shadows the unit counts most are the ones a room removes first. Difference and uniformity

The shadows a unit counts are the ones a room removes

ΔEITP's growth with display brightness is largest in the dark greys, and dark greys are where a lit room's light reflected off the screen sits. One candela a square metre of veiling luminance removes 22 per cent of the difference the unit gives a step at the bottom of the scale on a 1,000-candela display and nothing measurable at the top. The same veil raises the growth the unit reports across display levels from a factor of six to a factor of seventeen, because it destroys a dim display's shadows first.

 A halftone tint, between the two ideals and nearer one of them. For each pigment, how far the hue turns when it is tinted to the same lightness three ways: mixed with white pigment, mixed with white light, and printed as a halftone at a Yule–Nielsen factor of two. Right is towards a longer wavelength. 7 of the eight pigments' two ideals turn the hue opposite ways. The halftone mark sits between them at every pigment and between 26 and 44 per cent of the way from the light ideal to the paint one — so it keeps the additive tint's direction and loses about a third of its size. What the brain does

Printing does not change the sign

A tint mixed in paint and a tint mixed in light turn seven of eight pigments' hues opposite ways. A halftone sits between the two because light entering the paper between dots emerges under a dot, and the question was where. At the optical gain a coated sheet is fitted at, a halftone has travelled between 26 and 44 per cent of the way from the light ideal to the paint one — the same fraction for every ink — and only one pigment in eight changes direction.

 The same gradients, priced by a penalty and by a projection. Each row is one gradient held inside a coated press from six starts. The pale dots are the penalised relaxation — a free step, with a price for leaving the press — and the dark ones are the projected relaxation, which takes the free step and then moves each point to the nearest printable colour. Across is how much longer than the free path each result is, logarithmic. On the 10 gradients whose straight line leaves the press, the penalty leaves 9 starts at more than twice the free length and the projection leaves none. The projection's best route costs a median 0.02% against the penalty's 0.13%, and its spread across starts is 0.95% against 132.5%. Matching and measuring

A projection has no reason to detour

Holding a gradient inside a press by penalising the excursion turned the gamut's price from a number into a search: on five of ten crossing gradients some starting point leaves the relaxation trapped at more than twice the free length, and the spread across six starts runs to 271 per cent of the free path. Replacing the penalty with a projection — take the free step, then move each point to the nearest printable colour — leaves no trap on any gradient and a spread of 0.06 to 2.4 per cent.

 A four-ink inverse table's error along the grey axis, 9 nodes, black from L* 60. The colour difference between a grey asked for and the grey printed, from L 92 to 14, for a 9-node inverse table whose separation strategy begins black generation at L 60 (dashed line). Three tables: filled exactly from the strategy, mean 0.073; refitted with each node sliding along the strategy, 0.050; refitted freely, 0.036. To the left of the dashed line the strategy-bound refit lies close to the filled table; to the right it lies close to the free refit. Every table is exact at its nodes, marked on the axis. What it takes to deliver it

A strategy keeps the refit only where black prints

Refitting an inverse table's nodes against their own error halves it, and in a four-ink table the refit has to leave the separation strategy's choice of black alone. Held to the strategy, it keeps the whole of its gain wherever the strategy prints black and almost none of it where the strategy does not — and the share it keeps equals the share of the grey ramp printed with black, to within three hundredths. The start of black generation, predicted to be where the refit would fail, is where the table needs it least.

 The calibration error a narrow fourth channel survives, by where it is placed. For a red, green and blue ambient sensor with one more channel 10 nm wide, centred from 450 to 640 nm: the first calibration error at which some lamp of fourteen is misclassified. Dashed: the channel pooled into the whole residual, as the ambient sensor's disagreement was read; its best is ±1.50 per cent. Solid: the channel read on its own, as its departure from what the camera's white predicts for it; its best is ±4.0 per cent, at 450 nm. Crosses mark centres where the channel read alone does not separate the lamps at all. The two horizontal lines are the red, green and blue design (±1 per cent) and the design with a clear channel (±4 per cent). What a camera does

A narrow channel has to be read on its own

An ambient sensor with a clear channel tells structured lamps from smooth ones by reading deep red, and one without it reads structure but breaks at one per cent of calibration error. A narrow fourth channel between the camera's peaks was proposed to have both. Pooled into the sensor's disagreement with the camera, it reads structure and breaks at one and a quarter per cent — no better than the design it was meant to rescue. Read on its own, as its departure from what the camera predicts for it, a channel at 450 nm holds to four per cent, the clear channel's figure, and far red does not move it.

 Four ways to correct the corner, under each of four lamps, at 25°. The mean colour error left on the chart's coloured patches at the corner of the frame, under each lamp, after: the red row fitted under that lamp; one red row fitted on the chart under all four lamps with the grey held under daylight; a grey-card gain map calibrated under daylight; and the worst of the other lamps' rows used by mistake. daylight: 0.93, 1.33, 1.86, 4.91; tungsten: 0.99, 1.45, 1.87, 3.26; white LED: 0.36, 1.39, 3.28, 1.77; fluorescent: 0.37, 1.56, 3.09, 2.45. The pooled row holds every lamp between 1.33 and 1.56. What a camera does

One row for every lamp costs the lamps that lose least

A corner correction fitted under one lamp is right under that lamp and can be badly wrong under another, so a converter unsure of its lamp was offered a single row fitted under several at once. Pooled over four lamps, one row holds every lamp's corner between 1.33 and 1.56 colour differences — no lamp worse than a grey-card map, and none near the 4.91 a wrong row can leave. The price falls on the white LED and the fluorescent tube, which it leaves four times worse than their own rows, because they lose the least red at the corner and the pooled row is built to repair the lamps that lose the most. Two rows, one per class of lamp, keep almost all of both.

 Where chroma stops protecting a wall, in five rooms. The seventy-two paints in each of five rooms, sorted by how much light the wall returns and read in overlapping windows of eighteen: the rank correlation of the share of colour a satin finish takes with the wall's chroma, lightness held. Below zero a more saturated paint loses less; above, more. The dots are where each room's curve crosses: cube at 0.36, corridor at 0.37, low room at 0.38, one wall open at 0.29, two walls open at 0.24. Solid lines are closed rooms of three shapes; dashed are the cube with one and two walls opened. What a scene does

An open room hands over sooner

A gloss finish takes the least colour from a saturated dark wall and the most from a saturated pale one, and in a closed cube the sign changes at a wall returning about a third of the light. The prediction was that a less enclosed room would move that point up the lightness scale and that a room with a window would have no pale end. Opening one wall moves it down, from a luminance factor of 0.36 to 0.29, and opening a second to 0.24; stretching the room into a corridor or flattening it nudges it slightly up. The ambient does whiten, as predicted. A whiter ambient does not delay the colour term — it weakens it, so the other term takes over sooner.

 The share of its colour each display hue loses to a quarter of its luminance in white. The twenty-four most saturated colours of an sRGB display, one every fifteen degrees of HSV hue, each scaled to a luminance of 20 and given white of a quarter that luminance: the share of its colourfulness lost, in CIECAM16, CIELAB and Oklab. All three put the fastest loss at an orange and the slowest at a blue. At the orange end they are close — CIECAM16 27 per cent, CIELAB 28 per cent, Oklab 21 per cent — and at the blue end CIECAM16's 1.7 per cent is a quarter of the others'. What the brain does

White drains the blue last in every model

CIECAM16 says a dab of white costs a deep red wall four times the colour it costs a dark blue one, and the experiment proposed to test it asked whether observers lose colour in that hue order. Held at equal luminance and given equal doses of white, twenty-four display hues are ordered alike by CIECAM16, CIELAB and Oklab — an orange loses fastest and a blue slowest in all three. So the order cannot tell the models apart. What does is how much slower the blue is: CIECAM16 has it losing a sixteenth of what the orange loses, CIELAB and Oklab about a quarter. That ratio is the number an observer study has to measure.

 The appearance model's lightness-to-chroma balance, as the room takes a share of the adaptation. The median lightness pair's reading over the median chroma pair's, for twelve base colours, against the display's white, in a room of 20 cd/m². Solid: CAM16-UCS, with the viewer taking all, three quarters, half, a quarter and none of the adaptation from the display — darker lines take more from the display. Dashed: ΔEITP, which no room enters. With the display alone the model's balance falls 23 per cent; with half from the room, 13; with none from the display it does not move. Difference and uniformity

A lit room brings the units' medians together

As a display brightens, CAM16-UCS says chroma differences gain a quarter on lightness differences and ΔEITP says a tenth. All of the appearance model's movement comes from what the viewer is adapted to, and every calculation had the viewer adapted to the display alone. Give the room its share of the adaptation and the model's movement shrinks at every step: with a 20 cd/m² room supplying two thirds of it, the two units' medians fall by the same amount. What does not shrink is their disagreement about direction. The model still moves every colour the same way, ΔEITP still moves violets, reds and cyan-blues the other way, and in a lit room that becomes the whole of what separates them.

 How wrong a declared veil makes the unit, for four true veils. On a 100 cd/m² display, the worst error over grey steps from L 2 to L 90 — the size of the natural logarithm of the declared reading over the true one — against the veil declared, for rooms putting 0.1, 0.3, 1 and 3 cd/m² on the screen. Each curve reaches nought at its own true veil and rises on both sides. The flat stretch at the left is declaring almost nothing, which is declaring none: 0.22 for a true veil of 0.1, 0.54 for a true veil of 0.3, 1.14 for a true veil of 1, 1.89 for a true veil of 3. Difference and uniformity

A guessed veil halves the error

A colour difference that takes a display's absolute luminance leaves out the light a room reflects off the screen, and on an ordinary display in an ordinary room that makes it wrong about the darkest greys by a factor of three. Giving the unit the veil as a declared argument fixes that when the veil is known. The worry was that it never would be — that a guessed argument is no better than none. It is better: any declared veil up to about twice the true one beats declaring none, and one middling guess for every room halves the worst error. What a guess cannot do is reach ten per cent; that needs the veil known within a sixth, which is what a luminance meter aimed at a black screen gives.

 How far a mesopic match moves as the S weight opens: two rooms against one field. For every pair of the five lamps, the median over forty-two surfaces of how far a match moves as the rod signal's weight into the S channel goes from nothing to equal: pale for the two-room match, each half adapted to its own lamp; dark for a bipartite field whose two halves share one adaptation. The field's signal is larger for every pair, by ×1.5 to ×7.7. What the eye does

One field keeps what two rooms divide out

An asymmetric match can measure how strongly the rods feed the blue-yellow pathway, but set with the observer adapted to each lamp in turn it needs seventy-five settings on the best pair of lamps and hours of waiting between them. Putting the two lamps on the two halves of one field was proposed as the quick version, at the cost of a weaker signal. The signal is not weaker. Under one shared adaptation it is three times stronger for daylight against a white LED, and the best pair needs six settings. The adaptation that makes the slow version slow is also what was dividing the rods' contribution out of each half.

 How much of a coated press is left at each margin inside its boundary. The share of a coated press's printable volume in CIELAB that lies at least a given distance inside its boundary, for margins from half a unit to 32. A margin of one unit keeps 88%, two keep 80%, four 66% and eight 45%. At every margin up to 30 what is left is a single connected piece; at 31 units, with 0.14% of the volume left, it first splits, into a core of 623 cells and 2 fragments of one or two cells — the deepest point is 32.7 units inside, so what splits is the last crumb of the core, not a waist. Matching and measuring

A margin costs a press its corners

A gradient held inside a coated press by projection was predicted to tear if the press were first shrunk by a safety margin, at any pinch narrow enough for the shrinking to cut. The press has no such pinch: shrunk by any margin up to thirty CIELAB units it stays in one piece, and projected routes on it are neither trapped nor torn. What a margin costs is concentrated at the corners — the solid yellow moves nearly four times the margin to get inside, like the tip of a 31-degree spike.

 Narrowing the slit makes the declared-width bound looser, not tighter. The median ratio of the declared-width bound to the error it bounds, over sixty-eight notches under each lamp, with the slit and the grid narrowed together from five nanometres to one and each lamp declaring its true line width; the dotted lines are the bound with both variances known. Under the tube the declared bound goes from ×86 at five nanometres to ×253 at one, while the true-variance bound stays between ×12 and ×15. Under the projector the declared bound lies on its true-variance bound at every slit. What light is

A narrow slit shrinks the error, not the bound

A colour engine given two separately blurred spectral tables can bound its own error if the lamp declares its narrowest feature, and under a fluorescent tube that bound sits 86 times above the error at a five-nanometre slit. Narrowing the slit was supposed to close the gap, because a comb of narrow features stops fitting in a narrow window. It does the opposite. At one nanometre the error falls 27-fold and the bound only sevenfold, so the bound is 253 times loose and certifies no notch within a colour difference. The looseness was never the comb. It is the lamp's one declared peak, charged to every window.

 The ceiling a table reads for each window, against the lamp it came from. The fluorescent tube between 525 and 590 nanometres, with the green and yellow mercury lines, and for each window of a one- and a five-nanometre slit the highest the lamp could be there according to the table's own entries. The dashed line is the tube's peak, which the global declaration charges to every window. At one nanometre the ceiling follows the lamp, a median 2.05 times each window's true peak; at five it sits on the global peak for twenty nanometres either side of each line. What light is

A narrow table can vouch for its own peaks

The bound a colour engine can put on its bandpass error is loose because a lamp declares one peak for its whole spectrum and the bound charges it to every window. A table measured through a slit about as narrow as the lamp's lines can do better: no feature of the declared width can be blurred below a computable fraction of its height, so the table's own entries cap how tall the lamp can be in each window. At one nanometre that ceiling is never below the truth and sits twice above it, where the global peak sat twenty times above. The bound halves, and for the first time it certifies notches within a colour difference. At five nanometres it buys nothing at all.

 A finer grid shrinks both arrangements, and the gap between them stays. The mean error over sixty-eight notches of the best fixed pair of separate slits and of one slit on the reflected light at the grid's own step, on grids of one, two and five nanometres, under the tube and the projector. The tube's best pair falls from 0.256 to 0.019 and reaches the product's five-nanometre error of 0.016 without passing it; the product on the same fine grid is 0.0007. Under the projector the pair at one nanometre is 0.135. What light is

A finer grid keeps the price of separating

A sample's reflectance and a lamp's spectrum measured through separate slits can be reused under any other lamp; measured together through one slit on the reflected light, they are fifteen times more accurate and belong to that one lamp. A finer grid was expected to close the gap, because a one-nanometre table barely blurs anything. It closes most of the distance to the old answer and none of the distance to the new one. On a one-nanometre grid the best separate pair under a fluorescent tube is as good as one slit on the product used to be at five, and one slit on the product on the same fine grid is twenty-nine times better again.

 Every symmetric and every skewed fourth emitter, and the best of each at every cost. Each fourth emitter added to the three-emitter lamp, as its fidelity index against the number of the solid's directions a 40-nanometre slope limit costs more than a per cent: 315 symmetric bands and 1260 skewed ones over centre, widths and power. The two step lines are the best rendering each family reaches at or below each cost. They run a few directions apart and never further: rendering at 88 costs 103 directions with a symmetric band and 97 with a skewed one. The ringed point is the three emitters alone, Rf 78.6 at 62. Where the model breaks

A lopsided band does not break the trade

A fourth emitter added to a three-emitter LED renders colours better and costs a slow pigment its reach, and the position that renders best and the position that costs least were thirty-five nanometres apart. The obvious suspect was the shape: every emitter searched was a symmetric bump, and a band with its skirt on one side might fill a gap without splitting it. Freed to be lopsided, it saves between one and seven of the solid's three hundred directions. No band renders like the amber emitter and costs like the green one, the lopsided bands that help lean the other way from the proposal, and what the fourth band renders was never the lamp's white.

 The same lamps, charged by two kinds of slow pigment. Each lamp — the three emitters, a fourth band at seven positions, and the cheapest symmetric and lopsided bands rendering at 88 — as its fidelity index against the directions of the solid a slow pigment loses more than a per cent of its reach in, for a forty-nanometre slope limit (open) and for three absorption bands none narrower than forty nanometres (filled), joined by a line. Every line runs leftward: the band-shape pigment loses fewer directions under every lamp, and the lamps rendering best gain the most. At 88 the slope limit loses 50 and the band-shape pigment 33. Where the model breaks

The pigment decides the trade, not the band

A four-emitter lamp that renders well costs a slow pigment part of its reach, and reshaping the fourth band barely changes that. Changing the pigment changes it a great deal. Modelled as three absorption bands none narrower than forty nanometres, rather than as a forty-nanometre slope limit, a slow pigment loses a third fewer of the solid's directions under the best-rendering lamp, and the price of going from the three-emitter lamp to a fidelity of 88 halves. The lopsided band that helped a slope-limited pigment does nothing at all for this one. And at 590 nanometres a fourth band renders at 86 almost for free.

 Every lamp, charged by a slope limit and by bands in nanometres and in energy. For each lamp, ordered by fidelity index, the directions of the solid lost by a forty-nanometre slope limit, by three absorption bands none narrower than forty nanometres, and by three none narrower than 0.158 electronvolts. The energy convention loses fewest under every lamp: 20 against 23 and 31 under the three emitters, and 26 against 33 and 51 under the best-rendering lamp. Where the model breaks

Written in energy, the bands spare the green

A slow pigment modelled as absorption bands pays half what a slope limit pays for a four-emitter lamp's rendering, and a fourth band at 590 nanometres costs it almost nothing. That result assumed every band is at least forty nanometres wide, and a real absorption band has a width in energy, not in nanometres. Written that way it is narrower in the blue and wider in the red, and the fear was that the red would take back the amber. It does not. The band limit written in energy loses fewer directions under every lamp, the fourth band at 590 still costs six, and the price of rendering at 88 falls from ten directions to six — because the directions it wins back are green, not amber.

 The hue disagreement with the quantiser, and without it. For twelve base colours of CIELAB lightness 50 and chroma 40, the factor by which ΔEITP's reading of a lightness pair over its reading of a chroma pair changes from a 1.5 to a 10,000 candela white. With the perceptual quantiser, it runs from ×0.75 at the yellow-greens to ×1.23 at the violets. With the quantiser replaced by a pure power law and nothing else changed, it is ×1 at every hue. Difference and uniformity

The hue disagreement is the quantiser's

As a display brightens, the appearance model shifts every colour's balance between lightness and chroma by one factor, and ΔEITP shifts violets towards lightness and yellow-greens towards chroma. The proposed source was the perceptual quantiser acting on three cone-like signals at unequal levels. Replacing the quantiser with a pure power law and changing nothing else removes the disagreement exactly: every colour moves by a factor of one. With the quantiser, how far a colour moves follows the spread of its three signals, and which way follows where its S signal sits — above the other two, towards lightness; below, towards chroma.

 Three panels in three rooms: the veil alone, and the best single floor. On a 1,000 cd/m² display, for each panel and room, the worst error over five-per-cent grey steps from 0.005 cd/m² upwards of declaring only the room's veil (upper bar) and of the best single declared floor (lower bar). For an LCD without dimming and for an OLED the best floor is the display's black plus the veil, and it is exact. For the LCD it is also necessary: the veil alone misreads by 2.8 in a dark room. For a local-dimming LCD no single floor is exact, and in a dark room the best errs by 0.96. Difference and uniformity

A dimming panel has no single black

ΔEITP reads a display's shadows correctly only when it is told the floor under every pixel, and that floor has two parts: the room's light reflected off the screen and the display's own black. They fold into one declared number exactly, and on a bright LCD the display's part is the one that matters — leaving it out misreads the deepest shadows seventeenfold. A local-dimming panel breaks the folding, because its black changes with the picture: in a dark room its floors vary twenty-six times across one scene, and no single declaration comes within a factor of two and a half. A lit room squeezes them back into one number.

 CIECAM16's colourfulness loss by hue, with each suspect removed. The share of colourfulness each of the twenty-four most saturated sRGB colours loses when white of 25 per cent of its luminance is added, all at the same luminance, in CIECAM16 as published, with its eccentricity factor held at one, and with its cone-signal denominator replaced by its achromatic signal; CIELAB chroma for reference. The blue at 240 degrees loses 1.7 per cent as published and 1.6 with the hue factor removed, and 5.4 with the denominator replaced — the curve then lies close to CIELAB's. What the brain does

The blue resists through the model's denominator

CIECAM16 says a saturated blue is nearly impervious to added white: at equal luminance and an equal dose it loses a sixteenth of what an orange loses, where CIELAB and Oklab say a quarter. The model has two obvious places for that to come from — a hue factor fitted to make colourfulness equal across hues, and the sum of cone signals it divides its opponent signals by. Setting the hue factor to one changes nothing. Replacing the sum with the model's own achromatic signal makes the blue lose 0.233 of the orange's loss, beside CIELAB's 0.221 and Oklab's 0.223. The disagreement is one denominator.

308 essays on this thread, page 9 of 13 · all threads · all essays