Repository navigation
node blur
Home > Image & Fields > Blur
< Previous: Graph Mapper | Next: Combine >
Blur (node)Soften an image or a field over a disc of the given Radius; Strength mixes the result back with the original.
Category: Image & Fields | Add menu: Image & Fields > Blur
Blur averages what comes in over a disc around each point it is asked about. Softening an image before you sample it is the usual reason: a photograph full of grain makes a field full of grain, and every copy driven by that field jitters. One Blur in front of Image Sample settles the whole pattern down.
- In (any): the image, the field, or the plain number to soften.
- Radius (number): how far the averaging reaches, in millimeters of model space. Default 5.
- Strength (number): how much of the blurred result to keep, from 0 to 1. Default 1.
- Out (any): the softened result, the same kind of thing that went in.
In and Out are both typed any, and Blur hands back whatever it was given:
- An image (from the Image node) comes out as a blurred image, so it can still feed Image Sample, PlaceVariant, or anything else that takes an image.
- A value field comes out as a blurred field, ready for any field-aware input.
- A plain number blurs to itself. There is nothing around a single number to average with.
Radius is measured in millimeters on the canvas, not in pixels, so the same Radius softens by the same visible amount whatever resolution the image happens to be. For an image, Locus converts it to pixels using the image's own baseline, the two corner points you pinned it by, which means scaling the image on the canvas does not change how soft it looks.
Strength mixes the blurred result back over the original: 0 leaves the input untouched, 1 is fully blurred, and the values between fade from one to the other. It is the dial to reach for when a blur that fixes the grain also flattens detail you wanted to keep. A Radius of 0, or a Strength of 0, passes the input straight through.
The average is taken over a fixed disc: the center point plus three rings of samples, Gaussian-weighted so the middle counts most. The count is fixed, so blurring a lattice of hundreds of points costs the same per point no matter how large the Radius is.
Two things do not blur, by construction:
- Index-seeded fields. The Per point output of Random gives every copy its own draw by index rather than by position, so there is no neighborhood to average over and the field passes through unchanged.
- The preview thumbnail on the node. The small picture on the node face is a per-pixel lookup and cannot show a blur, so it keeps showing the source. The model ghost and everything downstream do show the blur.
< Previous: Graph Mapper | Next: Combine >
- Image: the picture this softens | Image Sample: turn the softened image into a field
- Graph Mapper: reshape a field's values rather than its detail
- Field-aware nodes: which inputs can read a field | The halftone pipeline