Image Similarity

Overview

Image Similarity node compares the current input frame against a reference image and reports a similarity score along with contours marking the regions that differ between the two.

Use this node to detect changes between frames, identify surface defects or contamination against a known-good reference, or monitor for unexpected scene alterations.

Input

Input Image

image required

The live frame to compare. Connect this to a camera or upstream image output.

Reference Frame

image

The reference image to compare against. When not connected, the node uses the first received frame as the reference.

Mode

string

Method used to compute the similarity score and difference map.

Values:

• ABS (default) — absolute pixel-level difference. Simple and fast; sensitive to brightness shifts.
• SSIM — Structural Similarity Index. Accounts for luminance, contrast, and structure; more perceptually meaningful than ABS.
• PSNR — Peak Signal-to-Noise Ratio. A logarithmic score where higher values indicate more similarity.

Threshold

integer required

Binary threshold applied to the difference map before contour extraction. Pixels with a difference value above this threshold are treated as changed regions.

Range: 0 to 255
Default: 128

Lower values make the node more sensitive to small differences; higher values suppress minor variations such as sensor noise.

Overlay Results

boolean required

When enabled, highlights the detected difference contours on the output frame. See Overlay Results.

Output

Overlay Image

image

Output frame from the node. If overlays are enabled, difference regions are highlighted on this frame.

Score

number

Similarity score based on the selected Mode. Higher scores indicate greater similarity for SSIM and PSNR; lower scores indicate greater similarity for ABS.

Difference Contours

array

Array of contour objects describing the pixel regions that differ between the input and reference frame. Each contour is a list of [x, y] points.