Module: Graph Based Segmentation (2D) ()

Description:

This is an experimental module from the Xtra Library: https://xtras.amira-avizo.com.

This Xtra implements efficient graph-based image segmentation by Felzenszwalb and Huttenlocher [1]. Note that this module performs 2D segmentation and if given a 3D image as input it performs an XY slice-by-slice segmentation. Users can use this form of segmentation to perform over-segmentation of grayscale images. MinSize parameter allows users to specify the minimum size in pixels that can be attributed to a single segmented label. Tweaking kParam as specified in [1] increases the scale of observations. If one requires a more 'regular' surface then kParam can be increased in value while lower values provide more 'irregular' separation.

This Xtra embeds DLIB library in C++ and DLIB is covered by a Boost License 1.0. The Xtra does not distribute the source for DLIB and is only distributed as a compiled DLL/binary. See http://dlib.net/license.html.

Citations:

[1] Felzenszwalb, Pedro F., and Daniel P. Huttenlocher. "Efficient graph-based image segmentation." International Journal of Computer Vision (IJCV), 2004.

Connections:

Data [required]
Input image. 8,16 or 32 bit images

Ports:

kParam

(default : 200). Scale of observations. Larger values provide more 'regular' separation and smaller values provide 'irregular' separation.

minSize

(in px)(default : 10). Filter out smaller noisy structures in a noisy image to get more regular regions