This example shows how to perform multiple segmentation in the image, using coarse and fine resolutions. We use the three methods for segmentation currently available in InterIMAGE, namely Region Growing, Baatz and Checkerboard segmentations. This InterIMAGE interpretation project can be downloaded here: download.

The image used in this example is the following:

Firstly, you should create a new project, called, for instance “multiresolution_segmentation”, with the images you are going to use in your project (in this case a single image will be used).

After that, create a structure to perform the segmentations. In the following example, we firstly created a node called “coarse_segmentation” which applies the Region Growing segmentation. We also created three child nodes, called “fine_segmentation”, “fine_checkerboard”, and “background” with Baatz, Checkerboard and “Dummy top down” algorithms respectively. The node “background” will just be associated to remaining elements, apart from the city.

The segmentation parameters used were:

• coarse_segmentation (TerraAIDA_Region_Growing_Segmenter)
  • Euclidean distance threshold: 40
  • Segments minimun size: 200
  • fine_segmentation (TerraAIDA_Baatz_Segmenter)
    • Compactness weight: 0.5
    • Color weight: 0.5
    • Scale parameter: 20
    • Decision rule:
      • mean('b1') < 230 AND mean('b2') < 230 AND mean('b3') < 230
  • fine_checkerboard (TerraAIDA_CheckerBoard)
    • Size of cells (pixels): 5
    • Decision rule:
      • ratio('b1') < 0.4
  • background (Dummy TopDown)

Then, start your project by clicking in the button highlighted in blue.

When the process finishes, the “Map Viewer” window will open and will show different levels and types of segmentation. According to the decision rules, the white and empty area will be associated to background, the roofs will be associated as well as segmented as a checkerboard in class fine_chequerboard and the city will be associated to the class fine_segmentation.