public class ThresholdConfidence extends ReduceElements<ObjectMask>
Combines object-masks by projecting the maximum confidence-level for each voxel and thresholding.

After thresholding, a connected-components algorithm splits the thresholded-mask into single-objects.

Depending on the number of objects, two types of projection occur:

  • the projection occurs globally on an identically sized image to the entire size (more efficient for a larger number of objects and/or small image), or
  • using an R-Tree clusters of intersecting boxes are found, and each is processed separately (more efficient for a smaller number of objects and/or large image).
Author:
Owen Feehan
  • Constructor Details

    • ThresholdConfidence

      public ThresholdConfidence(double minConfidence)
      Creates with a minimum-confidence level.
      Parameters:
      minConfidence - the minimum confidence of an element for its object-mask to be finally included (after merging).
    • ThresholdConfidence

      public ThresholdConfidence()
  • Method Details

    • reduce

      Description copied from class: ReduceElements
      Reduce a list of elements (each with a confidence score and a label) to a smaller-list.

      It is not guaranteed that the resulting list will have fewer elements than the input list, but never more.

      Specified by:
      reduce in class ReduceElements<ObjectMask>
      Parameters:
      elements - proposed bounding-boxes with scores.
      extent - an area in which all elements must exclusively reside.
      executionTimeRecorder - records the execution time of particular operations.
      Returns:
      the outcome of the operation, indicating which elements are retained / removed / added.
      Throws:
      OperationFailedException - if anything goes wrong.
    • getMinConfidence

      public double getMinConfidence()
      The minimum confidence of an element for its object-mask to be initially included for consideration (before merging).
    • setMinConfidence

      public void setMinConfidence(double minConfidence)
      The minimum confidence of an element for its object-mask to be initially included for consideration (before merging).
    • getMinNumberVoxels

      public int getMinNumberVoxels()
      The minimum number of voxels that must exist in a connected-component to be included.
    • setMinNumberVoxels

      public void setMinNumberVoxels(int minNumberVoxels)
      The minimum number of voxels that must exist in a connected-component to be included.
    • getThresholdNumberObjectsGlobal

      public int getThresholdNumberObjectsGlobal()
      When the number of objects is greater or equal than this, they are reduced globally, without separation. See class javadoc.
    • setThresholdNumberObjectsGlobal

      public void setThresholdNumberObjectsGlobal(int thresholdNumberObjectsGlobal)
      When the number of objects is greater or equal than this, they are reduced globally, without separation. See class javadoc.