Images
Image Convolution Workbench
Apply a bounded3×3,5×5 or7×7 numeric filter with explicit edge handling.
Loading tool…
How to use Image Convolution Workbench
Apply a bounded3×3,5×5 or7×7 numeric filter with explicit edge handling. An isolated pixel loop convolves RGB channels with your finite kernel, divisor and offset, while retaining each original alpha channel. Clamp, wrap and zero-padding edges are separate deliberate choices.
- Provide raster image, square kernel json, kernel divisor, channel offset.
- Choose the supported mode and run the tool. Review its assumptions and any reported errors.
- Copy the result or download a separate output; reset to clear the current form.
Good to know
- Static PNG, baseline JPEG or simple VP8/VP8L WebP; each≤8MiB,4,000,000 pixels and4,096px per side. Progressive JPEG, VP8X/animated WebP, APNG and compressed PNG profiles/text or PNG EXIF are excluded. Full container structure is checked before native decoding; this is not a complete codec validator. Browser decoding/export can still reject a file or change color profiles. Output is a fresh PNG without original container metadata.
- 1,048,576-pixel image; kernel coefficients magnitude≤1,000, absolute coefficient sum≤10,000; divisor magnitude1e−12–1e6. RGB is clamped0–255; alpha is preserved.
- 10-second worker startup then2-second computation limit. This is encoded-channel filtering, not color-managed physical blur or arbitrary shader execution.
- Inputs are processed on this device and are not automatically saved or uploaded.
Common questions
What happens at an image boundary?
Clamp repeats the nearest edge pixel, Wrap reads the opposite edge, and Zero supplies black outside the image. These produce different, explicitly testable filter results.
Does this change my original input?
No. The tool returns a separate result. Review that output before using it in another application.