replaces small connected components in a binary image image with background pixels.


replaces positive integers in a label matrix m with 0 if their tally is small.


replaces components consisting of n or fewer elements.

Details and Options

  • DeleteSmallComponents works with arbitrary 2D and 3D images or label matrices.
  • The following options can be given:
  • CornerNeighborsTruewhether to include corner neighbors
    MethodAutomaticmethod to determine a size threshold
  • Possible settings for the Method option include:
  • "Cluster"cluster variance maximization (Otsu's algorithm)
    "Mean"use the mean of component sizes as the threshold
    "Median"use the median of component sizes as the threshold
  • The default setting is Method->"Cluster", corresponding to cluster variance maximization.


open allclose all

Basic Examples  (3)

Delete small components in a binary image:

Delete small components in a 3D image:

Remove small components from a label matrix:

Scope  (2)

Remove components with fewer than 1000 pixels:

Delete small components from a label matrix:

Options  (2)

Method  (2)

Use the default cluster variance maximization method:

Delete components smaller than the mean component size:

Applications  (2)

Clean small foreground and background components:

Delete smaller components resulting from image segmentation:

Introduced in 2010
Updated in 2012