Creating Threshold Segmentations

If the features within an image have homogeneous pixel values, and if those values are significantly different from the background, then thresholding can be applied to create a region of interest within a range of data. Automatic thresholds can be computed using the Otsu threshold, or you can manually set the interval of values corresponding to the features within the image. In either case, the threshold will be displayed as a red highlight, as shown below.

Threshold segmentation of a fracture in a granite sample

Some image corrections may be required before applying a threshold for segmentation, such as inhomogeneous illumination of the image or intensity variations in the Z axis (see Image Filtering).
If required, you can refine the initial segmentation by processing (see Refining Threshold Segmentations), as well as separate connected components (see Separating Connected Components).
How to Create a Threshold Segmentation
  1. Define the intensity range that best describes the feature(s) that you need to segment (see Defining Intensity Ranges).
  2. Do one of the following:
    • Click the Add to New button in the Range box to create a new region of interest within the selected range.

      Information about the new region of interest is displayed in the lower section of the Data Properties and Settings panel (see ROI Properties and Settings).

    • Select an ROI on the Data Properties and Settings panel, or a class of a multi-ROI in the Classes and scalar information box on the Data Properties and Settings panel, and then click the Add button in the Range box.

  3. Refine the initial segmentation (see Refining Threshold Segmentations) or separate connected components (see Separating Connected Components).

An initial threshold segmentation may contain unwanted objects, such as mounting aids, stray pixels, and noise. You can do the following to refine threshold segmentations:

  • Apply the Process Islands options to remove unwanted objects (see Processing Islands). For example, you can isolate the largest object or objects in the region of interest by choosing Isolate nth First Biggest in the Process Islands sub-menu and then selecting the number of largest objects that you want to keep.

  • Edit the results by applying morphological operators such as Open, Close, and Smooth (see Applying Morphological Operations).

  • Fill inner areas (see Filling Inner Areas).

  • Use the ROI Painter tools to manually edit the segmentation (see ROI Painter).

An initial threshold segmentation may contain discrete objects that are grouped together. In this case, you can separate these connected components based on connectivity (see Connectivity) and then analyze and classify objects that share specific characteristics (see Measurements and Scalar Data for Multi-ROIs and Analyzing and Classifying Measurements).

Do any of the following to separate connected components in a threshold segmentation:

  • Right-click a region of interest in the Data Properties and Settings panel and then choose New Multi-ROI (6-Connected) or New Multi-ROI (26-Connected) in the pop-up menu. See 6-Connected and 26-Connected for an explanation of 6-connected versus 26-connected.

    A multi-ROI is created automatically in which each connected component is assigned to a class (see Multi-ROI Properties and Settings).

  • Right-click a class of a multi-ROI in the Classes and scalar information box on the Data Properties and Settings panel and then choose New Multi-ROI (6-Connected) or New Multi-ROI (26-Connected) in the pop-up menu. See 6-Connected and 26-Connected for an explanation of 6-connected versus 26-connected.

    The connected components in the multi-ROI are separated and appear as additional classes in the multi-ROI.