Remove a Solid-Colour Background from PNG
In your browser, without AI.
Edge-aware flood-fill · adjustable tolerance · 100% client-side
This tool replaces a target background colour in a PNG with full transparency, producing an RGBA image. It uses tolerance-based colour keying and optional edge-aware flood-fill — the same logic as Photoshop's Magic Wand combined with Select Similar — and runs entirely in your browser. It is built for logos, icons, scans, screenshots, and product photos shot against a uniform background. For hair, fur, or photographic backgrounds, an AI segmentation tool is the right choice; this one is not.
Upload your images
Drag & drop anywhere or click to browse properties.
Supports PNG, JPG, WebP, BMP, TIFF, SVG
Colour-key vs AI background remover — which one do you need?
Two completely different categories of tool live behind the phrase "remove background from PNG". They make opposite trade-offs. Pick the one that matches your image — not the loudest brand.
| Criterion | This tool — colour-key | AI segmenters (e.g. remove.bg) |
|---|---|---|
| Best for | Logos, icons, signatures, screenshots, scans, product shots on a flat colour | Photos with hair, fur, fabric edges, complex backgrounds |
| How it works | Pixel matching against a target colour + connected-component flood-fill | Neural-network semantic segmentation |
| Predictability | Deterministic — same input + tolerance = same output | Probabilistic — different model versions give different results |
| Privacy | Runs in your browser — image never leaves your device | Image uploaded to a server |
| Speed | ~1–2s for a 4000×4000 PNG on a mid-range laptop | 2–10s including network round-trip |
| Cost | Free, unlimited | Usually metered or watermarked above a free tier |
| Offline | Yes, after first load | No — requires connection |
| Edge accuracy on hair / fur | Poor — not its job | Excellent |
| Edge accuracy on hard edges | Excellent — pixel-exact | Often softens or rounds corners |
Rule of thumb: if the background is one colour and the edges are crisp, use this tool. If the subject has hair, fur, or sits in a real-world photo, use an AI segmenter.
How to remove a background — 4 steps
- 1
Upload your image
Upload a PNG, JPG, or WebP image (up to 25 MB). The file stays on your device — nothing is uploaded to any server.
- 2
Pick the background colour
Click the eyedropper and pick the background colour. Start with a tolerance of 10. Increase to 20–30 for shadows or JPEG noise. Toggle 'edge-aware' if your subject contains the same colour as the background.
- 3
Preview and adjust
Preview the transparent result over a checkerboard. If artefacts remain, adjust tolerance and re-run; if it looks clean, move on.
- 4
Download the transparent PNG
Download the result. Output preserves the original resolution and adds an RGBA alpha channel.
Stays on your device
Runs on the Canvas API — no upload, no server, no analytics on the file. Works offline after first load.
Any solid colour
Eyedropper or RGB/HEX input. Common presets: pure white, off-white scans, chroma green (#00B140), chroma blue (#0047BB).
Edge-aware
Flood-fill starts from the four image edges. The white inside the letter "O" or inside a product label is preserved automatically.
How the algorithm works
The tool reads your image into an off-screen canvas, walks the pixel buffer, and writes
alpha = 0 for every pixel whose colour distance
from the target falls within tolerance. There are two modes for selecting which pixels qualify.
Edge-aware (flood-fill)
Use when the subject contains the same colour as the background.
The algorithm seeds from the four image edges and removes only pixels connected to those seeds via colour-similar neighbours. Interior regions matching the background colour are preserved. This is what you want for:
- •Logos with white interiors (the holes inside O, A, B, P)
- •Product shots where the label has white text
- •Illustrations with enclosed white regions
Global replacement
Use when the background colour does not appear inside the subject.
Every pixel matching the target colour (within tolerance) becomes transparent, regardless of position. Faster and simpler — and the right choice for most chroma-key shots, isolated icons, and silhouette work.
Tolerance — the one knob you actually adjust
Real-world backgrounds are rarely flat. The tolerance slider sets the maximum colour distance (Euclidean RGB) between a pixel and the target before it is treated as background. Tolerance 0 matches only exact colours; 50+ swallows shadows, gradients, and JPEG ringing around edges. Push it too far and you start eating the subject's anti-aliased edge.
Which tolerance and mode should I pick?
A starting point — adjust by ±5 if you see halos or leftover background.
| Image type | Tolerance | Mode | Common pitfall |
|---|---|---|---|
| Pure-white studio PNG (logo, icon) | 5–10 | Edge-aware | Going too high erodes anti-aliased edges |
| Off-white paper scan / signature | 15–25 | Edge-aware | Paper texture under tolerance 10 leaves grey speckle |
| Product photo, white seamless | 20–30 | Edge-aware | Shadow under the product needs higher tolerance |
| JPG saved as PNG (compression noise) | 25–35 | Edge-aware | JPEG ringing creates halos at low tolerance |
| Chroma green screen (#00B140) | 30–40 | Global | Spill on subject — accept it or feather later |
| Chroma blue screen (#0047BB) | 30–40 | Global | Same as green — pick the colour with the eyedropper |
| Solid black background | 10–20 | Edge-aware | Dark subjects (text, shadows) get eaten |
| Gradient / photographic background | — | Wrong tool | Use an AI segmenter instead |
When this tool is the wrong choice
- ·Portraits or any photo with hair or fur — colour matching cannot resolve sub-pixel strands.
- ·Outdoor shots with sky, foliage, or any background that is not one colour.
- ·Semi-transparent objects (glass, smoke, water) — these need true matting, not keying.
- ·Subjects whose edge colour blends gradually into the background (motion blur, soft light).
For any of the above, an AI segmentation tool is the right category.
Troubleshooting
Parts of the background are still visible
Parts of the subject are being removed
Edges look jagged or have a visible halo
The green or blue screen isn't fully removed
The result has a faint coloured outline around the subject
The image processes slowly or crashes the tab
Why the output is PNG
Transparency requires an alpha channel. JPG and BMP have no alpha, so they cannot store a transparent background — the moment you save as JPG, the transparency collapses to white. WebP supports alpha and is smaller, but PNG is the universally compatible answer across email, Office, design tools, browsers, and operating systems. If you need WebP for the web, run the transparent PNG through a PNG-to-WebP converter as a second step.
Frequently asked questions
Can I remove any colour background?
Yes. Use the eyedropper or RGB/HEX input to target any solid colour — pure white, off-white scans, chroma green (#00B140), chroma blue (#0047BB), or anything else. Pixels matching that colour within the tolerance window are written as transparent.
Does it use AI?
No — and that's deliberate. AI segmenters are powerful for hair and fur but unpredictable on logos, signatures, and product photos with sharp edges. This tool uses exact colour matching and connected-component flood-fill, so the result is reproducible: same input plus same tolerance equals the same output. It also runs offline, with no upload and no rate limit.
Will it remove complex, photographic, or gradient backgrounds?
Not effectively. This tool targets uniform-colour backgrounds. For photographic, gradient, or hair-heavy backgrounds, segmentation requires AI — that is a different category of tool and a different trade-off.
Is the output really transparent?
Yes. The tool writes an RGBA PNG with alpha values of 0 for removed pixels. Open the file over a coloured background in any image viewer and there will be no white box — just your subject.
Why is the output PNG and not WebP or JPG?
JPG has no alpha channel at all, so transparency is impossible. WebP supports alpha and is smaller, but PNG has the broadest support across email, design tools, presentations, browsers, and operating systems. After removing the background, you can re-encode to WebP using our PNG-to-WebP converter.
Can I process multiple images at once?
Not as a batch — each image needs its own colour pick to stay precise. To convert many similar images sharing the same background colour, use the same hex value and tolerance and run them sequentially; the page is cached after first load, so each run is effectively instant.
Does it work offline?
Yes. After the first page load the tool is cached as a progressive web app and works without an internet connection.
Is there a file size or resolution limit?
Practical limit is what your browser tab can hold in memory — typically 4000×4000 to 8000×8000 pixels on a modern laptop. Files up to ~25 MB are safe. Very large images may take a few seconds because the work happens entirely on your CPU.