Why two phones read the same fabric differently

Two phones can give different readings for the same cloth because each camera has its own sensor and its own image processing, and neither sees colour exactly as the eye does. Shadecard removes most of that difference by correcting every capture against white paper, a grey card or a colour chart, but some always remains. For repeatable work, measure the target and its lab dips on one phone, with the same reference, in the same place.

Where the differences come from

A phone photo is not a measurement: it is the camera's interpretation of the scene.

• Automatic processing. Before you see a pixel, the phone applies auto white balance, auto exposure, tone curves, sharpening and noise reduction. Two phones make different decisions for the same scene.
• Different sensors. Camera sensors do not respond to light the way the human eye does, and every model differs. Two fabrics can look the same to one camera and different to another – a form of metamerism that no software can fully remove.
• The light itself. A card corrects the light's overall tint, but not every difference in its spectrum. Moving from window light to a tube light changes what any camera records.
• Angle and sheen. Tilt, pile direction and lustre change how much light the cloth reflects back into the lens.
• 8-bit JPEG. Deep shades such as navy and black have little signal, so small differences in processing show up as larger differences in the reading.

What Shadecard does about it

1. Lock. The app meters once on the fabric, then locks exposure and white balance so all three photos share one camera setting. It also sets anti-banding against mains flicker.
2. Measure in linear light. It undoes the image's gamma encoding before any maths, and averages the whole box instead of reading one pixel.
3. Correct against a known reference. White paper or a grey card sets the white balance and the lightness scale for that exact capture. The colour chart fits a full colour-correction matrix.
4. Report the error bar. The spread between the three photos is shown, and confidence is only high when the spread is under 1, a reference was used, and the lock held.

If a phone cannot hold the exposure lock, the result says This camera couldn't hold its exposure lock (AE/AWB lock)… and the confidence stays at medium. If still capture stalls, the reading comes from the lower-resolution preview stream and the result says so.

Getting consistent readings across a team

• One job, one phone. Measure the customer's target and every lab dip for it on the same phone. Differences between those readings then mean something.
• Same reference, same kind. Agree on one reference – ideally an 18% grey card – and use the same card each time. If you use paper, use the same brand.
• Same measuring spot. Pick one window or light-box position and always measure there, away from coloured walls.
• Check the spread. Compare readings only when both have a low spread and no serious HEADS-UP warnings.
• Use the colour chart for disputes. When two phones must agree as closely as possible, the 24-patch chart gives the fullest correction the app offers.

When a reading jumps on the same phone

If the same cloth gives two different readings a minute apart on one phone, the conditions changed. Look at both results' Engineering details (spread, reference, each shot) and HEADS-UP notes. The usual causes are a cloud passing the window, the phone's own shadow, a different tilt, or the paper moving out of the cloth's light.

Related

• Lighting and framing for a good reading
• White paper, grey card or colour chart?
• Blog: Metamerism: why lighting changes fabric colour