Image to Grayscale

What this tool is used for

• Converting an image for a monochrome print or a single-colour layout.
• Checking whether a design still reads without colour, which is a useful accessibility test.
• Producing a neutral version of a photo to sit behind text.
• Removing colour from a screenshot so an annotation stands out.
• Preparing an image for a context where colour would be distracting.

Frequently Asked Questions

Why not just average the RGB channels?

Because human vision is far more sensitive to green than to blue. A plain average makes greens too dark and blues too light; the standard luminance weights are 0.2126 R, 0.7152 G, 0.0722 B under Rec. 709.

Do the weights differ by standard?

Yes. Rec. 601 (used for older SDTV) uses 0.299/0.587/0.114, while Rec. 709 matches modern sRGB displays. The two give visibly different results for saturated colours, which is why greyscale conversions can disagree.

Should gamma be handled?

Properly, yes — sRGB values are gamma-encoded, so weighting them directly is mathematically wrong. Correct conversion linearises first, applies the weights, then re-encodes. Most fast implementations skip this and are slightly too dark.

Does it reduce the file size?

Often substantially, since colour information is discarded — PNG can store a true greyscale channel rather than three. A JPEG saved as greyscale drops its chroma planes entirely.

Is greyscale the same as desaturation?

Not quite. Desaturation in HSL sets saturation to zero, which uses lightness rather than luminance and looks flatter. Luminance-weighted greyscale preserves the perceived brightness relationships far better.

Why do two greyscale conversions of the same photo differ?

Because the channel weights differ between standards, and some tools convert in gamma-encoded space while others linearise first. Both are defensible, and they produce visibly different mid-tones.

Is greyscale a good accessibility check?

It is a useful one — removing colour reveals whether a design still communicates through contrast and layout alone, which approximates how it reads for some forms of colour blindness. It is not a substitute for a contrast ratio check.

Can colour be recovered afterwards?

No. Three channels are collapsed into one, so the mapping is many-to-one and irreversible. Keep the original — colourisation guesses plausible colour, it does not restore the real one.

Common errors and gotchas

• Assuming a single conversion formula. Different weightings produce visibly different results from the same source.
• Losing contrast between two colours that had similar luminance, which then become indistinguishable.
• Treating grayscale as smaller. The pixel data may reduce, but the format often stores three channels anyway.
• Converting a chart whose only distinction was colour, which makes the data unreadable.
• Re-saving as JPEG and finding the artefacts more visible than they were in colour.