Let colour support the larger image
Colour matching decides which tiny photograph sits in which cell. Get it right and the big picture appears effortlessly; push it too far and the tiles lose the colours that made them worth including.
Accessible technique disclosures
Average colour (RGB)
Each cell and tile becomes one average red-green-blue value; the closest tile wins. Simple and fast. Weak point: RGB distance doesn't match perception well, so some greens and skin tones swap oddly.
Perceptual matching in CIELAB
CIELAB, defined by the CIE in 1976, separates lightness from colour. Measuring difference as Delta E, especially with the CIEDE2000 formula, ranks tiles closer to how your eye would. Good for portraits.
Sub-cell matching
Splitting each cell into a 2 × 2 or 3 × 3 grid lets a tile with a bright corner land where a highlight falls. Edges and eyes sharpen noticeably; you'll need a bigger library to benefit.
Luminance-first matching
Matching brightness before hue keeps the structure of the image intact even with a limited palette. Useful for themed libraries that lack certain colours.
Tinting and overlay blending
Tiles are shifted towards the target colour, or the main image is laid over the mosaic at partial opacity. Effective rescue, but it is visible up close.
Why matching is a balancing act
Every mosaic negotiates between two honesties. One is honesty to the main image: the viewer at a distance should see a convincing portrait or landscape. The other is honesty to the tiles: up close, each photograph should still look like itself, with its real sky and real skin. A library with perfect colour coverage satisfies both. Real libraries don't, so the matching method and the amount of correction decide which side gives way.
Reading a tint slider
Most makers express correction as a percentage. As a starting point:
- 0–10%: tiles look untouched; the big image relies entirely on library coverage.
- 10–25%: the sweet spot for many personal projects; gentle shifts close gaps without obvious colour casts.
- 25–40%: the main image reads strongly, but white tiles may turn coral or olive and faces in tiles take on odd hues.
- Above 40%: the tiles become texture; the work reads as a filtered photograph more than a mosaic.
There's no wrong setting, only an intention. A memorial mosaic where every tile is a cherished photo may deserve low tint and a bigger library. A bold event banner seen from ten metres can take more.
Matching for different subjects
| Main image | Suggested method | Tint starting point | Watch for |
|---|---|---|---|
| Close-up portrait | CIELAB with sub-cell | 15–20% | Eyes losing definition |
| Sunset landscape | Average colour, palette-first library | 10–20% | Banding in the sky gradient |
| Logo or lettering | Luminance-first | 25–35% | Ragged edges on curves |
| Pet photo, fur texture | Sub-cell 3 × 3 | 15% | Noisy, busy texture |
Prepare the main image first
Small edits to the main image often do more than any slider. Increase contrast slightly so shadows and highlights separate; simplify a cluttered background with a blur or a darker tone; crop tighter so the subject occupies more cells. Some makers let you mark areas of interest, such as a face, for smaller tiles or stricter matching. Use it if available.
Smooth gradients without banding
Clear skies and studio backdrops show stepped bands when the library has gaps in a hue range. Error diffusion, the idea behind Floyd–Steinberg dithering, spreads a cell's colour error to its neighbours so the eye averages the steps away. A few tools offer a similar option; otherwise, adding more tiles in that hue is the reliable cure.
Once colour behaves, plan the physical result: prepare a mosaic that reads at scale.