Gaussian Blur Redaction

A mathematical smoothing technique that uses a Gaussian function to reduce image detail and obscure sensitive visual information.

Gaussian Blur Redaction works by convolving an image with a specific kernel (a 2D matrix of weights) to average pixel intensities across a defined radius. Operators typically adjust the sigma value (the standard deviation of the Gaussian distribution) to control the level of data suppression. While effective for aesthetic privacy in media (like blurring faces or license plates), it remains susceptible to deconvolution attacks. Security researchers have demonstrated that AI models can often reconstruct original text from blurred regions if the kernel parameters are known. For high-stakes privacy (such as PII on government documents), solid masking is often preferred over this mathematical filtering method.

https://en.wikipedia.org/wiki/Gaussian_blur