Edge Detection

The Edge Detection operator is a real-time image-processing filter that highlights the boundaries between objects, textures, and shapes in a video frame. It works by looking at how quickly pixel intensity changes from one neighbour to the next — the steeper the change, the more likely the pixel sits on an edge — and renders those edges as bright lines on a dark background.

Because the analysis happens per pixel on the GPU, the operator runs at full broadcast frame rates and can be applied to live sources, file playback, or the output of an upstream operator chain.

How it works

Edge detection uses a gradient kernel that samples each pixel together with its immediate neighbours and asks "how different are they?" Where the answer is "very different" (a face against a sky, a high-contrast caption, the silhouette of a building), the operator outputs a bright pixel; where neighbours are similar (a flat wall, a single-tone background), it outputs a dark one.

The operator offers several classic algorithms, each with its own tradeoff between noise immunity, edge sharpness, and processing cost:

• Sobel — the default; a balanced choice with good noise immunity and clean edges on most footage.
• Prewitt — similar in spirit to Sobel with simpler weighting; slightly softer edges.
• Laplacian — a single-kernel operator that responds to edges in all directions at once. Faster, but more sensitive to noise.
• Roberts Cross — a small, fast 2×2 kernel that excels on clean, high-contrast material; can look noisier on flatter footage.
• Scharr — an enhanced Sobel with better rotational symmetry; useful when fine edge orientation matters.

If you don't know which one to pick, start with Sobel and only switch if the result looks too soft (try Roberts Cross or Scharr) or too noisy (stay with Sobel and tighten the thresholds described below).

Output modes

Detecting the edges is half the operator; the other half is choosing how to show them. The Output mode setting offers five styles:

• Edges Only — pure white edges on a black background. The classic look, and the right starting point for stylised effects or for piping the result into a downstream operator.
• Edges Coloured — the same edges drawn in a chosen colour rather than white. Lets you tint outlines to match a brand palette or scene mood without an extra colour-correction pass.
• Overlay — edges drawn directly on top of the original picture, leaving the rest of the frame untouched. Good for blueprint-style technical visualisations or reveal effects where the source content stays visible underneath.
• Edge Alpha — the original RGB image with the edge strength used as the alpha channel. Hands the matte directly to compositing, with the source colours preserved.
• Edges as Alpha — the original RGB image where the edges define transparency. Useful for cut-out style effects where the outlined regions punch through to a background.

Tuning the result

Three controls fine-tune which differences register as edges and how strongly they show:

• Minimum threshold — drops anything weaker than this. Raise it to suppress sensor noise and only keep strong outlines; lower it to reveal fine detail at the cost of more graininess.
• Maximum threshold — pixels above this register at full intensity. Edges between the minimum and maximum fade smoothly, giving anti-aliased outlines instead of hard on/off lines.
• Edge strength — a sensitivity multiplier. Use values above 100% when edges are subtle (low-contrast footage); below 100% when the picture is already very edgy.

A Reset command in the panel returns these three to their defaults (5 %, 50 %, 100 %) — handy when you've tuned yourself into a corner and want a clean starting point.

When to use Edge Detection

• Stylised video effects — sketch-like, neon-outline, technical-blueprint, or comic-book looks for creator workflows.
• Mask generation — feed the edge map into a downstream keyer or compositor to build alpha channels driven by image structure rather than colour.
• Motion graphics — overlay dynamic edge tracings on top of footage for music videos, broadcast IDs, or sting transitions.
• Technical analysis — visualise object boundaries when reviewing footage for QC, machine-vision pre-work, or scientific capture.
• Rotoscoping assistance — surface the edges that a manual rotoscope artist would otherwise have to find by eye, speeding up frame-by-frame masking.

Edge Detection - Settings

Algorithm Settings

Algorithm Settings — pick the edge detection method and how the result is presented.

Edge Thresholds

Edge Thresholds — control which differences register as edges and how strongly they show.

Edge Color

Edge Color — colour of edges when using a coloured-edges or overlay output mode.

Overlay Settings

Overlay Settings — controls used only when OutputMode is set to Overlay.

Additional Options

Additional Options — extra creative controls for edge appearance.

Performance and Properties

Performance and Properties — status messages from the operator.

Inherits from: AbstractOperator, AbstractAudioMetering.

See also: Edge Detection in Script Engine Objects.