Can the original pixels be recovered from a blur?

Sometimes, yes. Research (notably "Defeating Image Obfuscation with Deep Learning", Hill et al. 2016) has shown that low-radius Gaussian blur and coarse pixelation can be reversed for known-alphabet text (e.g. blurred numbers, names) using ML trained on the original font. For text that absolutely must not be recoverable, use the BLACK BAR option — it overwrites the pixels entirely.

What blur radius does the tool use?

The Gaussian blur uses a radius of ~12 pixels by default. That's strong enough to make most photographic content unrecognizable to humans, but as noted above, ML can sometimes reverse it on text. Increase the radius (or use black bars) for higher-stakes content.

Is the original image uploaded anywhere?

No. Everything happens via Canvas 2D in your browser. Open DevTools → Network during processing and you'll see zero outbound requests. The censored image is downloaded directly from your browser's memory.

Why PNG and not JPEG?

PNG is lossless, so the unmodified parts of the image stay pixel-perfect. JPEG compression would soften the entire image (not just the censored regions), which is often undesirable when you're redacting a screenshot.

Can I move or resize a box after I've drawn it?

Yes — click any existing box to select it, then drag its corners or center. Hit Delete to remove. The original image underneath is preserved until you export, so you can iterate freely.

Will the EXIF metadata (location, camera info) survive?

No. Canvas 2D drops EXIF when re-encoding to PNG. That's actually a privacy feature in this context — your photo's GPS coordinates and camera serial don't carry over into the censored output.

File & Format Converters · Free tool

Image Blur & Censor

Blur or black-out parts of an image, like faces or license plates, by dragging rectangles and exporting as PNG. A free, instant tool that runs in your browser with no signup.

Updated June 2026

Found this useful?Email Buy Me a Coffee

What it does

Drop a screenshot, photo, or document scan, and drag rectangles over the parts you want to hide — faces, real names, email addresses, credit-card numbers, license-plate digits, anything sensitive. For each box, pick a censor style: Gaussian blur (soft, looks natural in photos), mosaic / pixelate (the standard "TV-news pixelation" pattern), or solid black bar (when you want to guarantee the original is unrecoverable).

The tool is designed for the everyday privacy task — sharing a screenshot in a public Slack channel or Twitter thread without exposing identifiers, redacting receipts before submitting an expense report, anonymizing student work before sharing it as an example. Output is a PNG that preserves the rest of the image at original quality.

Important security note: blur and pixelate are visually obscuring but not cryptographically unrecoverable — research has shown that low-radius blur and coarse pixelation can sometimes be reversed by ML models trained on the original alphabet (see "Defeating image obfuscation with deep learning", Hill et al. 2016). For truly sensitive data (passwords, full credit-card numbers, SSNs), use the solid black bar option — it overwrites the pixels entirely, making recovery impossible.

All processing runs locally via the Canvas 2D API. The image never leaves your browser, and no metadata about your boxes or the source is logged anywhere.

Embed this tool on your siteShow snippet

Paste this snippet into any page. Loads on-demand (lazy), no tracking scripts, and sized to most dashboards. Replace the height to fit your layout.

<iframe src="https://freetoolarena.com/embed/image-blur-censor" width="100%" height="720" frameborder="0" loading="lazy" title="Image Blur & Censor" style="border:1px solid #e2e8f0;border-radius:12px;max-width:720px;"></iframe>

Embed docs →

Example input & output

Input

Screenshot of a Slack thread with 3 names and 2 emails visible

Output

Same screenshot with 5 black bars over the names and emails — the rest of the message is fully readable.

Black bars are the safest choice for text data because the underlying pixels are completely overwritten. Blur and pixelate look softer but in theory leave residual signal.

How to use it

Drag-drop your image into the upload area or click to browse.
Drag a box over the area you want to censor — release the mouse to commit. You can add as many boxes as you need.
For each box, pick a censor style: Blur (soft, photographic), Pixelate (TV-news mosaic), or Black bar (solid, unrecoverable).
Click Save / Download to export a PNG with the censoring baked in. The output is the original resolution.
If a box is wrong, click it and either re-drag the corners or hit Delete to remove it. The original photo underneath is preserved until you export.

When to use this tool

Sharing a screenshot publicly that contains real names, email addresses, internal links, or coworker faces.
Anonymizing student/intern work or onboarding screenshots before posting them externally.
Redacting receipts, invoices, or shipping labels before sharing for support tickets.
Quick-redact photos before posting on social media (license plates, house numbers, neighbor faces in a backyard shot).

When not to use it

When you need cryptographically unrecoverable redaction of structured data (passwords, full PAN credit cards, SSNs) — use the BLACK BAR option, not blur/pixelate. ML can sometimes reverse low-strength blur on known-format text.
PDF documents — this is image-only. For PDFs, use a PDF redaction tool (most professional tools have one).
Video — you'd need to mask each frame; use a video editor with motion-tracked masks.
When the document has been fingerprinted with steganography or watermarks — censoring a region doesn't erase those.

Frequently asked questions

Can the original pixels be recovered from a blur?: Sometimes, yes. Research (notably "Defeating Image Obfuscation with Deep Learning", Hill et al. 2016) has shown that low-radius Gaussian blur and coarse pixelation can be reversed for known-alphabet text (e.g. blurred numbers, names) using ML trained on the original font. For text that absolutely must not be recoverable, use the BLACK BAR option — it overwrites the pixels entirely.
What blur radius does the tool use?: The Gaussian blur uses a radius of ~12 pixels by default. That's strong enough to make most photographic content unrecognizable to humans, but as noted above, ML can sometimes reverse it on text. Increase the radius (or use black bars) for higher-stakes content.
Is the original image uploaded anywhere?: No. Everything happens via Canvas 2D in your browser. Open DevTools → Network during processing and you'll see zero outbound requests. The censored image is downloaded directly from your browser's memory.
Why PNG and not JPEG?: PNG is lossless, so the unmodified parts of the image stay pixel-perfect. JPEG compression would soften the entire image (not just the censored regions), which is often undesirable when you're redacting a screenshot.
Can I move or resize a box after I've drawn it?: Yes — click any existing box to select it, then drag its corners or center. Hit Delete to remove. The original image underneath is preserved until you export, so you can iterate freely.
Will the EXIF metadata (location, camera info) survive?: No. Canvas 2D drops EXIF when re-encoding to PNG. That's actually a privacy feature in this context — your photo's GPS coordinates and camera serial don't carry over into the censored output.

Learn more

Explore more file & format converters tools

100% in-browserNo downloadsNo sign-upMalware-freeHow we keep this safe →