pgraphbasedsegmentation

Performs pixel classification on image using boundary information between regions.

Synopsis

pgraphbasedsegmentation sigma k minimum-region-area [-m mask] [im_in|-] [rg_out|-]

Description

The pgraphbasedsegmentation exploits boundary between regions to segment color images. It adopts a graph-representation of the region and uses inter- and intra-region intensity homogeneity to determine boundaries. The intra-region intensity within a region is defined as the largest edge weight of the minimum spanning tree built from this region, and the inter-region intensity is defined as the minimum edge weight that connects these two regions.

The result is the region map rg_out.

Parameters

sigma is used to smooth the input image before segmenting it.
k is the value for the threshold function. It controls the fineness level of segmentation.
minimum-region-area specifies the minimum allowable region area (in pixels) contained in the segmented image. It's an integer greater than zero given in pixels.

Inputs

im_in: a 2D color image.

Outputs

rg_out: a region map.

Result

Returns the number of regions or FAILURE.

Examples

Segments the tangram pieces:

  pgraphbasedsegmentation 0.5 500 20 examples/tangram.pan a.pan

C++ prototype

Errc PGraphBasedSegmentation( const Imc2duc &ims, Reg2d &rgd, float sigma, float threshold, int minimumRegionArea );

Reference

Pedro F. Felzenszwalb and Daniel P. Huttenlocher, "Efficient Graph-Based Image Segmentation", International Journal of Computer Vision, 59(2) September 2004.