Denoise Image (OpenCV)

@opencv1 VersionEnhance

Classical CPU image denoiser over an Image: applies non-local-means, bilateral, Gaussian, or median filtering and returns a cleaned Image in the same color space. No model, no GPU. Pick this for sensor noise, JPEG quantisation, or salt-and-pepper impulse.

How it fits

Typical backends

Noisy capture cleaned before OCR: median or non-local-means denoising removes speckle so character recognition reads cleanly.

JPEG-degraded archive cleaned before object detection so quantisation artefacts do not trip the detector.

Web-served frame cleaned before classification to stabilise label confidence under sensor noise.

Caveats

FallbackAn unrecognised method silently falls back to non-local-means rather than erroring.
Parameter interactionstrength maps to a different physical quantity per method (filter strength, colour / spatial sigma, or blur sigma), and higher values remove more noise while blurring more detail.
Parameter interactionkernel is clamped to at least 1 and rounded up to the next odd value, is ignored by the non-local-means path, and is treated as a minimum diameter of 5 by the bilateral path.
Parameter interactioncolor only affects non-local-means and is ignored on images with fewer than three channels.
I/O contractThe output keeps the input color space; an empty Image is returned unchanged.
State lifetimeAll settings are captured once at startup; runtime changes have no effect and require a redeploy.

Versions

75b237a2defaultlatest
Automated release
6/21/2026