Houdini 20.0 Nodes Geometry nodes

Texture Feature geometry node

Uses computer vision to identify feature points in an image.

On this page


This node finds “prominent” (contrasty) features (such as corners) in an image, where the image is represented as a 2D volume, where voxel values are scalar (brightness) or vector (color).

You can restrict the search area by wiring into the mask input. The volume(s) in the mask input must be the exact same size as the volume(s) in the first input.

How to

To...Do this

Convert an image file to a 2D volume

If you have an image file and need to convert it into a 2D volume for use with this node, use the File node. Wire the output of the File node into this node.

If you import a multi-image format such as OpenEXR, the File node will convert each subimage/AOV into a separate volume. You can use the Group field of the Texture Feature node to operate on one specific volume (for example, @name=C).

Identify features in an image

  1. Set the Maximum feature count to the the number of features you want to identify in the image.

  2. If the node identifies too many features, you can limit the feature count, or lower the Quality tolerance. If the node identifies too few features, you can increase the Quality tolerance.

  3. To prevent dense clustering of features, increase the Minimum spacing.

Turn off feature labels in the viewport

When this node is selected and you choose the Handles tool (Enter) in the viewport, the operation toolbar has a Label features menu to control whether to draw labels on the identified features.



A list of volume primitives to use from the first input. If this is blank, the node uses all volumes in the input.

Maximum Feature Count

The maximum number of feature points to find.

Minimum Spacing

The minimum distance to maintain between feature points, in voxels.

Quality Tolerance

Controls how “prominent” (contrasty) features must be to appear in the output. Lower this number to get fewer, more prominent features. Increase this to get more, less prominent features.

Blur Radius

Increasing this blurs the input image more before looking for features. Increasing this may help reduce “false positives” in noisy images. Decreasing it may help identify features in overly blurry images.

Gradient Radius

Increasing this increases the number of neighboring pixels the node looks at to compute changes in value in the input image. Increasing this may help reduce “false positives” in noisy images. Decreasing it may help identify features in overly blurry images.


The evaluation method used to score features.

Corner Detection

Uses the Harris detection algorithm to find corner points.

Minimum Eigen Value

Uses Eigen Values from a local covariance matrix to find corner points.

Corner Weight

When Method is Corner Detection, this controls how much weight to give to apparent corners (points where areas of different values meet).


Whether to output the “features” as voxels in a sparse volume, or 3D points.

Output Name

When Output is Volume, use this as the volume name.


Volume Image

A two-dimensional volume to analyze for features.

Volume Mask

Optional. Only look for features in the firs input volume in voxels that have a corresponding non-zero value in this volume. The two volumes must be the same size.


TextureFeatureBasicExample Example for Texture Feature geometry node

This example demonstrates how to use the Texture Feature node with some example image data.

See also

Geometry nodes