Denoising

← Back to Image Preprocessing

• Gaussian smoothing: fast and simple; blurs both noise and edges. Appropriate when spatial resolution is not critical.

• Median filtering: preserves edges while removing impulsive (salt-and-pepper) noise.

• Bilateral filtering: edge-preserving smoothing; useful when structural boundaries must be maintained.

• Non-local means (NLM): exploits self-similarity across the image; particularly effective for Gaussian noise in microscopy images at the cost of higher computation.

See also

• Spatial Filtering — the underlying convolution-kernel theory behind these filters

• Contrast Enhancement — usually applied after denoising

• Image Preprocessing — every other preprocessing operation