Example: Shape Features

Functions > Image Processing > Feature Extraction > Example: Shape Features

Use the shape_features and concomp functions to compute shape-related features of an image and perform connected component labeling of pixels.

shape_features is useful for computing features such as centroids and spatial variances and moments of an input image. This function is used with image morphology and connected component labelling techniques to classify objects in an image. The function returns a matrix whose rows correspond to the component labels in the image, and whose columns correspond to calculated features. You can check which columns correspond to which feature by checking the overview of Shape Features.

concomp performs connected component labeling of pixels with grayscale value fg in matrix M.

For information on using this example, refer to About Image Processing Examples.

1. Read in an image.

(part.gif)

2. Define the structuring elements matrix (4 orthogonal neighbors) and its row/column coordinates.

3. Turn the image into a simple, connected component image by binarizing, using morphology functions to simplify it.

<region id="ID0E1S1Z" top="816" left="24" width="409.42" height="21.2" xmlns="http://schemas.mathsoft.com/worksheet50">
  <math>
    <define xmlns="http://schemas.mathsoft.com/math50">
      <id label="None">Mclean</id>
      <apply>
        <id>erode</id>
        <sequence>
          <apply>
            <scale />
            <real>255</real>
            <apply>
              <id>dilate</id>
              <sequence>
                <apply>
                  <scale />
                  <real>255</real>
                  <apply>
                    <id>binarize</id>
                    <id>M</id>
                  </apply>
                </apply>
                <id>Melem</id>
                <id>r</id>
                <id>c</id>
                <id>b</id>
              </sequence>
            </apply>
          </apply>
          <id>Melem</id>
          <id>r</id>
          <id>c</id>
          <id>b</id>
        </sequence>
      </apply>
    </define>
    <resultFormat>
      <decimal precision="3" show-trailing-zeros="false" radix="dec" imaginary-value="i" />
      <matrix size="12,12" offset="0,0" />
    </resultFormat>
  </math>
</region>

Erosion spreads light objects in an image further apart. Dilation enlarges light objects, reassembling an image that has become disjointed. The intensity threshold at which to erode or dilate is set to 1.

4. Apply the connected component labeling function to the simplified image, using 8-connected neighbors.