Class SoLocalMaximaQuantification

  • All Implemented Interfaces:
    SafeDisposable

    public class SoLocalMaximaQuantification
    extends SoImageVizEngine
    SoLocalMaximaQuantification engine. The SoLocalMaximaQuantification engine extracts the local maxima of an image using a neighborhood analysis.

    Input_Image[i,j] is a local maximum if the 2 conditions below apply:

    1. Considering k in [1,p], l in [1,q] and m in [-p,p], InputImage(i,j) verifies:
    InputImage( i, j) > InputImage(i + m, j - l)
    InputImage( i, j) > InputImage(i - k, j)
    InputImage(i, j) >= InputImage(i + k, j)
    InputImage(i, j) >= InputImage(i + m, j + l)
    2. Input_Image[i,j] is higher than a specified threshold
    InputImage(i, j) >= threshold

    Graphic visualization of the neighborhood analysis

    Examples:

    Example 1:
    Input parameters: p=0, q=0, threshold=4, number=50.
    The local maxima will be the points higher or equal to 4.

    Number of local maxima detected: 7
    Positions of the local maxima:
    (7,3) (6,5) (5,7) (6,7) (6,8) (2,9) (6,9)
    This corresponds to a threshold.

    Example 2: Input parameters:p=1, q=1, threshold=3, number=50.
    Number of local maxima detected: 4
    Positions of the local maxima: (7,3) (6,5) (5,7) (2,9)

    Remark: If there is an area in the image at constant level, the engine returns the upper left corner of this area. The other points of the area are not considered as �local maxima�.

    Example 3: Input parameters: p=2, q=2, threshold=3, number=50.
    Number of local maxima detected: 2
    Positions of the local maxima: (7,3) (5,7)

    Remark: The algorithm is not performed on the image border depending on n and m values. That is the reason why the point (2,9) is not considered as a local maximum.

    Local Maximum Sub-Pixel Position Determination:
    On example 3, the algorithm detects 2 local maxima. The positions are converted to sub-pixel positions:

    X Sub-pixel Position Determined by Parabola Approximation

    Y Sub-pixel Position Determined by Parabola Approximation

    File format/default:

    LocalMaximaQuantification {

      computeMode MODE_AUTO
      inGrayImage NULL
      localWindow 3 3 3
      thresholdLevel 0.0f
      precisionMode PIXEL
      numberOfPatterns 1
    }


    Library references: maxextract maxextract3d