ComfyUI Radial Average Amplitude Spectrum
A ComfyUI custom node that computes and visualizes the Radial Average Amplitude Spectrum of an input image — a 1D frequency-domain analysis tool useful for characterizing image texture, noise, and spatial frequency content.
What It Does
- Converts the input image to grayscale using luminance weights.
- Applies a 2D FFT and shifts the DC component to the center.
- Computes the amplitude spectrum (
|FFT|). - Averages the amplitude values at each integer radial distance from the center, producing a 1D spectrum.
- Renders the spectrum as a log-log plot and returns it as a standard ComfyUI
IMAGEtensor.
Installation
Option A — Copy into custom_nodes
cp -r ComfyUI_RadialAvgAmpSpectrum /path/to/ComfyUI/custom_nodes/
Option B — Clone directly
cd /path/to/ComfyUI/custom_nodes
git clone https://github.com/your-repo/ComfyUI_RadialAvgAmpSpectrum.git
Then restart ComfyUI. The node will appear under the analysis category.
Dependencies
All dependencies are part of the standard Python scientific stack that ComfyUI already ships with:
| Package | Purpose |
|---|---|
numpy |
FFT and array operations |
matplotlib |
Plot rendering |
Pillow |
Image conversion |
torch |
Tensor I/O (ComfyUI standard) |
No additional pip install steps are required in a typical ComfyUI environment.
Node Reference
Node name: Radial Avg Amplitude Spectrum
Category: analysis
Inputs
| Name | Type | Description |
|---|---|---|
image |
IMAGE |
Input image (any resolution, RGB or grayscale) |
plot_width |
INT |
Width of the output plot image in pixels (default: 512) |
plot_height |
INT |
Height of the output plot image in pixels (default: 512) |
line_color |
STRING |
Matplotlib color string for the spectrum line (default: red) |
log_x |
BOOLEAN |
Use logarithmic scale on the frequency axis (default: true) |
log_y |
BOOLEAN |
Use logarithmic scale on the amplitude axis (default: true) |
Outputs
| Name | Type | Description |
|---|---|---|
spectrum_image |
IMAGE |
Log-log plot of the 1D radial average amplitude spectrum |
Interpreting the Output
- X axis (frequency): Spatial frequency in cycles per pixel, expressed as radial distance from the DC center (1 = lowest non-DC frequency, Nyquist ≈ min(H,W)/2).
- Y axis (amplitude): Mean amplitude of the FFT at that radial frequency bin.
- A steep negative slope indicates a smooth, low-frequency-dominant image.
- A flat or slowly decaying spectrum indicates high-frequency content (sharp edges, noise, fine texture).
License
MIT