11.8. Simple Linear Iterative Clustering (SLIC)

11.8.1. 概観

This filter creates superpixels based on k-means clustering.

Superpixels are small cluster of pixels that share similar properties. Superpixels simplifies images with a great number of pixels making them more easy to be treated in many domains (computer vision, pattern recognition and machine intelligence). GIMP's aim is more humble: create a posterization effect.

k-means clustering is one of the most used algorithms to create superpixels. Superpixel color is the mean of pixels color in the corresponding region.

11.8.2. フィルターの呼び出し方

This filter is found in the image window menu under FiltersArtisticSimple Linear Iterative Clustering….

11.8.3. オプション

図17.205 Simple Linear Iterative Clustering options

「Simple Linear Iterative Clustering」 options

Presets, Preview, Split view
[注記] 注記

These options are described in 「Common Features」.

Regions size

Increasing regions size collects more pixels, and so superpixels size increases also.

図17.206 Regions size example

「Regions size」 example

Regions size = 16

「Regions size」 example

Regions size = 32


Compactness

Superpixels borders may be irregular. Increasing this option gives superpixels more regular border.

図17.207 Compactness example

「Compactness」 example

Compactness = 20

「Compactness」 example

Compactness = 40: look at the dome.


Iterations

How many times filter is repeated. Increasing this value gives more details.

図17.208 Regions size example

「Regions size」 example

Iterations = 1 (default)

「Regions size」 example

Iterations = 15


Clipping

The result of this filter can be larger than the original image. With the default Adjust option, the layer will be automatically resized as necessary when the filter is applied. With the Clip option the result will be clipped to the layer boundary.