Check a reduced-detail image effect

Choose a visual effect with a clear purpose

Image Blur and Pixelate applies a reduced-detail effect to a whole image or a selected rectangle. It can soften a distracting background or create an intentionally blocky graphic. It should not be treated as certified redaction. A face, address or account number can remain recognizable from context, partially covered edges or a weak effect. When information must be removed, replace it with an opaque covering in a suitable editor and independently inspect the flattened export. This page does not identify faces, recognize text or decide which parts of an image are sensitive. You choose the region and evaluate the result.

Worked example: compare a small checkerboard

Create an owned image containing alternating dark and light squares with a bright border. Upload it and leave the effect on the full image. Pixelate should turn the source into larger blocks as strength increases, while the blur mode smooths the reduced-detail reconstruction. The canvas dimensions should remain the same as the decoded source dimensions. Download a PNG and open it separately to compare the border and interior pattern. This example makes the effect easy to inspect without using a real identity document or somebody else's face. It also shows that strength describes an image operation rather than a measured privacy score.

Select only the intended rectangle

Drag across the preview to define a rectangle. The effect then applies to that rectangle while the rest of the source remains unchanged. Dragging in the opposite direction should still create a rectangle between the two chosen corners. Coordinates are mapped from the displayed canvas back to source pixels, so inspect the selection after using a small screen or resizing the page. The preview uses pointer interaction for mouse and touch input. If a selection is too small to process meaningfully, it may leave little visible change. Clear the selection to return to a whole-image effect before starting a different comparison.

Strength is relative to the region

The implementation downsamples the chosen area, then reconstructs it with smoothing enabled for blur or disabled for pixelation. It is not a calibrated Gaussian blur with a guaranteed radius. The same strength value can look different on a large photograph and a small icon because their pixel dimensions differ. A selected region also provides less material to downsample than an entire frame. Increase strength while watching the actual subject, then inspect the exported file at its native size. A thumbnail that looks unreadable can reveal recognizable details when enlarged or placed beside an unmodified copy of the source.

Formats and transparency

PNG is a useful first choice when you need a lossless raster comparison or transparent areas. JPEG usually suits photographic output but cannot preserve transparency; transparent pixels are placed against a white background for that export. WebP availability and encoding behavior depend on the browser. The page checks whether the requested format was actually encoded rather than silently giving a PNG a different extension. All exports are flattened raster images without editable effect layers. Keep an original copy if you may need another crop, stronger effect or different format. Repeated lossy exports can accumulate artifacts without improving the usefulness of the image.

Know what the exported pixels omit

Canvas export creates a new raster from the decoded image. It does not preserve the original file's full metadata, animation, color-management workflow or application-specific editing history. The tool accepts bounded static raster formats, not an arbitrary image archive or a multi-page document. An animated source is outside the intended workflow. Use a dedicated editor when you need to preserve layers or inspect every frame. Re-encoding may reduce some metadata, but this page does not certify that all personal information is absent. Visible surroundings, filenames and information embedded in the pictured scene require separate attention.

Recover from invalid images and stale previews

A file extension alone does not establish that its bytes can be decoded. Corrupt, empty or oversized input should produce a visible error and should not leave a previous picture available for download as though it belonged to the new file. The page bounds file size, dimensions and total pixels to avoid obvious memory overload, although device capacity still varies. Wait for reading to finish before changing parameters. If a supported image cannot be decoded, reopen it in a trusted local editor and save a fresh static copy. Renaming a document to PNG does not convert its contents.

Processing and disclosure

File reading, image decoding, region selection and raster export take place in the browser. This feature does not upload an image to a blur-processing service. Ordinary page assets, analytics and advertising requests are separate from that processing and are described in the site's privacy information. A local workflow does not protect against a compromised device, clipboard capture or somebody viewing the screen. Delete unnecessary downloaded copies when finished with a sensitive task. For a support example, use a checkerboard or fictional label instead of the real image. The page does not retain an authenticated project history that you can recover after closing the tab.

A practical acceptance checklist

Use a known source with distinct corner colors and an interior pattern. Confirm that whole-image processing changes the interior, a selected rectangle leaves the outside intact and clearing the selection returns to the full frame. Test a corrupt file and verify that export becomes unavailable. Download PNG, then another supported format, and inspect their actual dimensions and signatures rather than relying on the filename. Finally, view the result at normal reading size and enlarged. These checks establish the observed image effect. They do not demonstrate irreversible de-identification, legal anonymization or a guarantee that hidden information cannot be inferred.

Image Blur & Pixelate — Whole Image or Selected Region