Gaussian Parameters

Configure the Gaussian filter with sigma and kernel size. Sigma controls how much the image is blurred; kernel size sets the window the Gaussian is evaluated in.

Sigma (Standard Deviation)

Sigma is the actual blur control — it sets the spread of the Gaussian weights and therefore how strongly the image is smoothed.

• Range: 0.5 - 5.0

• Low values (0.5-1.0): Subtle smoothing, preserves more detail

• Medium values (1.5-2.5): Balanced noise reduction

• High values (3.0-5.0): Strong smoothing, may blur edges

Kernel Size

Kernel size sets the width of the window (in pixels, odd) that the Gaussian is sampled over. Internally it is converted to scipy's truncate value so the filter spans exactly that many pixels; the weights outside the window are cut off and the remainder is renormalised. It does not, by itself, set the blur strength — sigma does — but it caps how far the blur can reach.

• Available: 3, 5 (default), 7, 9, or 11 pixels

• Small windows (3x3): Cut off the tails of the Gaussian. Combined with a large sigma this behaves like a box-limited blur rather than a true Gaussian.

• Larger windows (7x7 to 11x11): Let a wide Gaussian extend fully. Keep the window at least about four times sigma to avoid truncating a smooth Gaussian.

Tip: Start with sigma=1.5 and kernel=5x5, then adjust based on results. Increase sigma if noise remains; decrease if details are lost. Widen the kernel only when you also raise sigma.