Fabric Value Sorter
Averaging red, green and blue misranks one fabric pair in four. Yellow has 13x the luminance of blue; averaging says twice.
| Colour | Lightness | Band | Averaging says |
|---|
Value is not the average of red, green and blue
Every quilting book says to sort by value, and everyone squints at the pile trying to do it by eye. The reason it is hard is that the eye is not equally sensitive to the three primaries: green carries about 72% of perceived brightness, red 21% and blue only 7%.
| Colour | Averaging says | Actual lightness | |
|---|---|---|---|
| Red | 33.3 | 53.2 | |
| Yellow | 66.7 | 97.1 | |
| Green | 33.3 | 87.7 | |
| Cyan | 66.7 | 91.1 | |
| Blue | 33.3 | 32.3 | |
| Magenta | 66.7 | 60.3 |
A pure yellow has 12.9 times the luminance of a pure blue. Averaging puts them a factor of two apart. It is not a rough approximation of value — it is a different quantity that happens to agree sometimes.
One pair in four comes out wrong
Swept over a grid of 216 colours — 23,220 pairs — the naive average puts 20% in the wrong order. Count the pairs it calls exactly equal when they are not, which is also a failure to rank them, and it gets 24.7% wrong. Roughly one pair in four.
And the worst case is not a near miss. Pure green and the blue-violet rgb(51, 0, 204) average to exactly the same number, so averaging calls them identical in value. Their real lightness is 87.7 against 26.9 — a gap of sixty L* units, most of the entire scale. In a black-and-white photograph one is nearly white and the other nearly black.
The two colours on the left average identically. The two on the right are what they actually look like in greyscale.
Which is exactly why the standard advice is to photograph the pile in black and white, or squint through a red value filter. Both do the weighting for you. This tool does the same arithmetic and shows you which two fabrics you are most likely to get backwards.
Why the bands are not at a third and two thirds
Luminance is how much light comes off the cloth; lightness is how light it looks, and the two are related by a cube root. Splitting light, medium and dark evenly on the perceptual scale puts the boundaries at 7.7% and 36.2% of the light — nowhere near a third and two thirds. It is the same reason a photographer's middle grey reflects about 18% of the light rather than 50%, and the same reason a fabric that seems "mid tone" is usually darker than you think.
Notes
- Sample the fabric, not the picture of it. Photograph your pile in flat even light and pick the colours off that. A studio shot on a retailer's site has been adjusted, and a fabric that reads medium in the shop can read dark at home.
- Value is relative. A fabric is only "dark" next to something lighter. The bands here are absolute so you can compare across projects, but what matters in a quilt is the spacing between the fabrics you actually chose.
- A quilt needs all three bands. The commonest reason a quilt reads flat despite lovely fabrics is that they are all mediums. The tool flags a missing band.
- Print and texture average out at a distance. A busy print reads as its average value from across the room, which is often lighter than the darkest thing in it. Squint at it — that is exactly what squinting does.
How to use
- Photograph your fabric pile in flat, even light.
- Pick the hex colours off the photo and paste them in.
- Switch to greyscale to see the pile the way value works.
- Check for a missing band — that is what makes a quilt read flat.
Frequently asked questions
What does "value does the work, colour gets the credit" mean?
That the light-and-dark structure of a quilt is what makes the pattern readable from across the room, and the hues are decoration on top of it. A quilt whose fabrics are all the same value reads as a flat rectangle however beautiful the colours are.
Why is sorting fabric by value so hard by eye?
Because the eye is not equally sensitive to the three primaries — green carries about 72% of perceived brightness, red 21% and blue only 7%. So a pure yellow has nearly thirteen times the luminance of a pure blue, and no amount of staring will make that intuitive.
Can I just average the red, green and blue?
No, and it fails more often than you would guess. Across a grid of 216 colours the naive average puts 20% of pairs in the wrong order, and counting the pairs it calls exactly equal when they are not, it gets 24.7% wrong. Roughly one pair in four.
How badly can averaging get it wrong?
Completely. Pure green and the blue-violet rgb(51, 0, 204) average to exactly the same number, so averaging calls them identical in value. Their true lightness is 87.7 against 26.9 — a gap of sixty points, most of the whole scale. In a black-and-white photograph one is nearly white and the other nearly black.
Why do people photograph fabric in black and white?
Because a greyscale conversion does the weighting for you. A red value filter does the same thing optically. This tool does the same arithmetic and additionally tells you which two of your fabrics you are most likely to get backwards.
Why are the light and dark bands not at a third and two thirds?
Because how much light a fabric reflects and how light it looks are related by a cube root. Splitting evenly on the perceptual scale puts the boundaries at 7.7% and 36.2% of the light — the same reason a photographer's middle grey reflects about 18% rather than 50%.
How much value range does a quilt need?
Enough that the bands are genuinely distinct — as a rough guide, at least thirty points between your lightest and darkest, with something in all three bands. The commonest reason a quilt disappoints is that everything in it turned out to be a medium.
Does a busy print count as one value?
From a distance, yes — a print reads as its average value from across the room, which is usually lighter than the darkest thing in it. Squinting is exactly the right test, because squinting is a low-pass filter and so is distance.
Does this send anything anywhere?
No. Everything is calculated in your browser and nothing is stored or uploaded.
🔒 This tool runs entirely in your browser. Nothing you enter is uploaded, logged, or stored.