Module imagingbook.common
Package imagingbook.common.sift

Class SiftParameters

java.lang.Object
imagingbook.common.sift.SiftParameters
All Implemented Interfaces:
ParameterBundle<SiftDetector>
public class SiftParameters extends Object implements ParameterBundle<SiftDetector>
Default parameters; a (usually modified) instance of this class may be passed to the constructor of SiftDetector.

  • Field Summary

    Fields
    Modifier and Type
    Field
    Description
    boolean
    DEBUG
    Set true to output debug information.
    int
    nAngl
    Number of angular descriptor bins.
    SiftDetector.NeighborhoodType3D
    nhType
    Type of neigborhood used for peak detection in 3D scale space.
    int
    nOrient
    Number of orientation bins in the feature descriptor (angular resolution).
    int
    nRefine
    Maximum number of iterations for refining the position of a key point.
    int
    nSmooth
    Number of smoothing steps applied to the orientation histogram.
    int
    nSpat
    Number of spatial descriptor bins along each x/y axis.
    int
    P
    Number of octaves in Gaussian/DoG scale space.
    int
    Q
    Scale steps (levels) per octave.
    double
    rhoMax
    Maximum principal curvature ratio used to eliminate line-like structures (3..10).
    double
    sDesc
    Spatial size factor of descriptor (relative to feature scale).
    double
    sFscale
    Scale factor for converting normalized features to byte values in [0,255].
    double
    sigma0
    Base scale at level 0 (base smoothing).
    double
    sigmaS
    Sampling scale (nominal smoothing level of the input image).
    double
    tDomOr
    Minimum value in orientation histogram for dominant orientations (rel.
    double
    tExtrm
    Minimum difference to all neighbors in DoG peak detection (max.
    double
    tFclip
    Maximum value in normalized feature vector (0.2 recommended by Lowe).
    double
    tMag
    Minimum magnitude required in DoG peak detection (abs.
    double
    tPeak
    Minimum DoG magnitude required for extrapolated peaks (abs.

  • Constructor Summary

    Constructors
    Constructor
    Description
    SiftParameters()
     

  • Method Summary

    Methods inherited from class java.lang.Object

    clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

    Methods inherited from interface imagingbook.common.util.ParameterBundle

    getValidParameterFields, printToStream, printToString, validate

  • Field Details

    • DEBUG

      public boolean DEBUG
      Set true to output debug information.

    • nhType

      public SiftDetector.NeighborhoodType3D nhType
      Type of neigborhood used for peak detection in 3D scale space.

    • sigmaS

      public double sigmaS
      Sampling scale (nominal smoothing level of the input image).

    • sigma0

      public double sigma0
      Base scale at level 0 (base smoothing).

    • P

      public int P
      Number of octaves in Gaussian/DoG scale space.

    • Q

      public int Q
      Scale steps (levels) per octave.

    • tMag

      public double tMag
      Minimum magnitude required in DoG peak detection (abs. value).

    • tPeak

      public double tPeak
      Minimum DoG magnitude required for extrapolated peaks (abs. value).

    • tExtrm

      public double tExtrm
      Minimum difference to all neighbors in DoG peak detection (max. 0.0005).

    • nRefine

      public int nRefine
      Maximum number of iterations for refining the position of a key point.

    • rhoMax

      public double rhoMax
      Maximum principal curvature ratio used to eliminate line-like structures (3..10).

    • nOrient

      public int nOrient
      Number of orientation bins in the feature descriptor (angular resolution).

    • nSmooth

      public int nSmooth
      Number of smoothing steps applied to the orientation histogram.

    • tDomOr

      public double tDomOr
      Minimum value in orientation histogram for dominant orientations (rel. to max. entry).

    • nSpat

      public int nSpat
      Number of spatial descriptor bins along each x/y axis.

    • nAngl

      public int nAngl
      Number of angular descriptor bins.

    • tFclip

      public double tFclip
      Maximum value in normalized feature vector (0.2 recommended by Lowe).

    • sFscale

      public double sFscale
      Scale factor for converting normalized features to byte values in [0,255].

    • sDesc

      public double sDesc
      Spatial size factor of descriptor (relative to feature scale).

  • Constructor Details