🌀 2D Image FFT

Source pattern Frequency filter Filter radius 16 Band width 8 Show phase (not magnitude) Image size

64 × 64

128 × 128

Image size64 × 64

Dominant freq—

Energy retained—

Compute time—

A 2D FFT splits an image into spatial frequencies. The centre of the spectrum (after fftshift) is the DC term (average brightness); spikes off-centre are periodic patterns. Low-pass keeps the centre → blur. High-pass removes the centre → edges. A notch deletes a bright spike → removes repeating stripes while keeping the rest. We run a real radix-2 FFT on rows then columns (separable transform).

Spatial image

Magnitude spectrum (log)

Filtered reconstruction