ee.Algorithms.Image.Segmentation.SNIC

AI-generated Key Takeaways

• The SNIC algorithm performs superpixel clustering on an input image.

• The output includes a band of cluster IDs and the average values for each input band within those clusters.

• If seeds are not provided, the output will also contain a seeds band indicating the generated seed locations.

• The size and compactness parameters significantly influence the clustering results and may need adjustment based on the image.

Superpixel clustering based on SNIC (Simple Non-Iterative Clustering). Outputs a band of cluster IDs and the per-cluster averages for each of the input bands. If the 'seeds' image isn't provided as input, the output will include a 'seeds' band containing the generated seed locations. See: Achanta, Radhakrishna and Susstrunk, Sabine, 'Superpixels and Polygons using Simple Non-Iterative Clustering', CVPR, 2017.

Usage	Returns
ee.Algorithms.Image.Segmentation.SNIC(image, size, compactness, connectivity, neighborhoodSize, seeds)	Image

Argument	Type	Details
image	Image	The input image for clustering.
size	Integer, default: 5	The superpixel seed location spacing, in pixels. If 'seeds' image is provided, no grid is produced.
compactness	Float, default: 1	Compactness factor. Larger values cause clusters to be more compact (square). Setting this to 0 disables spatial distance weighting.
connectivity	Integer, default: 8	Connectivity. Either 4 or 8.
neighborhoodSize	Integer, default: null	Tile neighborhood size (to avoid tile boundary artifacts). Defaults to 2 * size.
seeds	Image, default: null	If provided, any non-zero valued pixels are used as seed locations. Pixels that touch (as specified by 'connectivity') are considered to belong to the same cluster.

Examples

Code Editor (JavaScript)

// Note that the compactness and size parameters can have a significant impact
// on the result. They must be adjusted to meet image-specific characteristics
// and patterns, typically through trial. Pixel scale (map zoom level) is also
// important to consider. When exploring interactively through map tile
// visualization, the segmentation result it dependent on zoom level. If you
// need to evaluate the result at a specific scale, call .reproject() on the
// result, but do so with caution because it overrides the default scaling
// behavior that makes tile computation fast and efficient.


// Load a NAIP image for a neighborhood in Las Vegas.
var naip = ee.Image('USDA/NAIP/DOQQ/m_3611554_sw_11_1_20170613');

// Apply the SNIC algorithm to the image.
var snic = ee.Algorithms.Image.Segmentation.SNIC({
  image: naip,
  size: 30,
  compactness: 0.1,
  connectivity: 8,
});

// Display the original NAIP image as RGB.
// Lock map zoom to maintain the desired scale of the segmentation computation.
Map.setLocked(false, 18, 18);
Map.setCenter(-115.32053, 36.182016, 18);
Map.addLayer(naip, null, 'NAIP RGB');

// Display the clusters.
Map.addLayer(snic.randomVisualizer(), null, 'Clusters');

// Display the RGB cluster means.
var visParams = {
  bands: ['R_mean', 'G_mean', 'B_mean'],
  min: 0,
  max: 255
};
Map.addLayer(snic, visParams, 'RGB cluster means');

Python setup

See the Python Environment page for information on the Python API and using geemap for interactive development.

import ee
import geemap.core as geemap

Colab (Python)

# Note that the compactness and size parameters can have a significant impact
# on the result. They must be adjusted to meet image-specific characteristics
# and patterns, typically through trial. Pixel scale (map zoom level) is also
# important to consider. When exploring interactively through map tile
# visualization, the segmentation result it dependent on zoom level. If you
# need to evaluate the result at a specific scale, call .reproject() on the
# result, but do so with caution because it overrides the default scaling
# behavior that makes tile computation fast and efficient.


# Load a NAIP image for a neighborhood in Las Vegas.
naip = ee.Image('USDA/NAIP/DOQQ/m_3611554_sw_11_1_20170613')

# Apply the SNIC algorithm to the image.
snic = ee.Algorithms.Image.Segmentation.SNIC(
    image=naip, size=30, compactness=0.1, connectivity=8
)

# Display the original NAIP image as RGB.
m = geemap.Map()
m.set_center(-115.32053, 36.182016, 18)
m.add_layer(naip, None, 'NAIP RGB')

# Display the clusters.
m.add_layer(snic.randomVisualizer(), None, 'Clusters')

# Display the RGB cluster means.
vis_params = {'bands': ['R_mean', 'G_mean', 'B_mean'], 'min': 0, 'max': 255}
m.add_layer(snic, vis_params, 'RGB cluster means')
m