GenericRegionMerging

Segmentation using region merging algorithm

Description

This application allows to use the Generic Region Merging library (GRM).
It provides currently 3 homogeneity criteria: Euclidean Distance, Full Lambda Schedule and Baatz & Schape criterion.

Parameters

Input Image -in image Mandatory

Ouput Label Image -out image [dtype] Mandatory

Homogeneity criterion to use -criterion [bs|ed|fls] Default value: bs

  • Baatz & Schape
  • Euclidean Distance
  • Full Lambda Schedule

Threshold for the criterion -threshold float Mandatory

Number of iterations -niter int Default value: 0

Activate it to boost the segmentation speed -speed int Default value: 0

Weight for the spectral homogeneity -cw float Default value: 0.5

Weight for the spatial homogeneity -sw float Default value: 0.5

Examples

From the command-line:

otbcli_GenericRegionMerging -in QB_Toulouse_Ortho_XS.tif -out labels.tif -criterion bs -threshold 60 -cw 0.7 -sw 0.3

From Python:

import otbApplication

app = otbApplication.Registry.CreateApplication("GenericRegionMerging")

app.SetParameterString("in", "QB_Toulouse_Ortho_XS.tif")
app.SetParameterString("out", "labels.tif")
app.SetParameterString("criterion","bs")
app.SetParameterFloat("threshold", 60)
app.SetParameterFloat("cw", 0.7)
app.SetParameterFloat("sw", 0.3)

app.ExecuteAndWriteOutput()

Limitations

This application is not adapted for large images.