Package gov.nih.mipav.model.algorithms

Class AlgorithmMeanShiftClustering

java.lang.Object
java.lang.Thread
gov.nih.mipav.model.algorithms.AlgorithmBase
gov.nih.mipav.model.algorithms.AlgorithmMeanShiftClustering
All Implemented Interfaces:
ActionListener, WindowListener, Runnable, EventListener
public class AlgorithmMeanShiftClustering extends AlgorithmBase
The java code is ported from C++ code downloaded from http://coewww.rutgers.edu/riul/research/code.html. The relevant web page section says: Adaptive mean shift based clustering C++ code implementing an (approximate) mean shift procedure with variable bandwith (in high dimensions). The algorithm is described in Mean shift based clustering in high dimensions: A texture classification example. For comments, please contact Bogdan Georgescu or Ilan Shimshoni. The original paper was written by Bogdan Georgescu, Ilan Shimshoni, and P. Meer in the proceedings of ICCV 2003. The original files were fams.cpp and fams.h. This code was ported by William Gandler.

  • Field Details

    • EVERY_POINT

      public static final int EVERY_POINT
      See Also:

    • SELECT_ON_JUMP

      public static final int SELECT_ON_JUMP
      See Also:

    • SELECT_PERCENT

      public static final int SELECT_PERCENT
      See Also:

    • FAMS_FKL_NEL

      private static final int FAMS_FKL_NEL
      See Also:

    • FAMS_FKL_TIMES

      private static final int FAMS_FKL_TIMES
      See Also:

    • FAMS_MAX_K

      private static final int FAMS_MAX_K
      See Also:

    • FAMS_MAX_L

      private static final int FAMS_MAX_L
      See Also:

    • FAMS_BLOCKSIZE

      private static final int FAMS_BLOCKSIZE
      See Also:

    • FAMS_BLOCKSIZE2

      private static final int FAMS_BLOCKSIZE2
      See Also:

    • FAMS_MAXITER

      private static final int FAMS_MAXITER
      See Also:

    • FAMS_ALPHA

      private static final double FAMS_ALPHA
      See Also:

    • FAMS_PRUNE_WINDOW

      private static final int FAMS_PRUNE_WINDOW
      See Also:

    • FAMS_PRUNE_MINN

      private static final int FAMS_PRUNE_MINN
      See Also:

    • FAMS_PRUNE_MAXM

      private static final int FAMS_PRUNE_MAXM
      See Also:

    • FAMS_PRUNE_MAXP

      private static final int FAMS_PRUNE_MAXP
      See Also:

    • FAMS_PRUNE_HDIV

      private static final int FAMS_PRUNE_HDIV
      See Also:

    • Bs

      private static final int Bs
      See Also:

    • Bs2

      private static final int Bs2
      See Also:

    • FAMS_FLOAT_SHIFT

      private static float FAMS_FLOAT_SHIFT

    • K

      private int K

    • L

      private int L

    • k_neigh

      private int k_neigh

    • data_file_name

      private String data_file_name

    • input_directory

      private String input_directory

    • choosePoints

      private int choosePoints

    • jump

      private int jump

    • percent

      private double percent

    • fixedWidth

      private boolean fixedWidth

    • width

      private float width

    • findOptimalKL

      private boolean findOptimalKL

    • epsilon

      private float epsilon

    • Kmin

      private int Kmin

    • Kjump

      private int Kjump

    • FAMS_DO_SPEEDUP

      private boolean FAMS_DO_SPEEDUP

    • nPoints

      private int nPoints

    • nDims

      private int nDims

    • pttemp

      private float[] pttemp

    • modesImage

      private ModelImage modesImage

    • prunedModesImage

      private ModelImage prunedModesImage

    • noLSH

      private boolean noLSH

    • raFile

      private RandomAccessFile raFile

    • byteIntBuffer

      private final byte[] byteIntBuffer
      byte array for int *

    • byteFloatBuffer

      private final byte[] byteFloatBuffer
      byte array for float *

    • endianess

      private boolean endianess

    • array1

      private int[] array1

    • minVal

      private float minVal

    • maxVal

      private float maxVal

    • points

      private AlgorithmMeanShiftClustering.famsPoint[] points

    • data

      private int[][] data

    • dataSize

      private int dataSize

    • rr

      double[] rr

    • psel

      private int[] psel

    • nsel

      private int nsel

    • modes

      private int[][] modes

    • hmodes

      private int[][] hmodes

    • npm

      private int npm

    • prunedmodes

      private int[][] prunedmodes

    • nprunedmodes

      private int[] nprunedmodes

    • M

      private int M

    • M2

      private int M2

    • hashCoeffs

      private int[] hashCoeffs

    • K_

      private int K_

    • L_

      private int L_

    • t_cut_res

      private int[][] t_cut_res

    • t_old_cut_res

      private int[][] t_old_cut_res

    • t_old_m

      private int[] t_old_m

    • t_m

      private int[] t_m

    • t_m2

      private int[] t_m2

    • t_hjump

      private int[] t_hjump

    • nnres1

      private int nnres1

    • nnres2

      private int nnres2

    • tt1

      private long tt1

    • srand

      private Random srand

  • Constructor Details

    • AlgorithmMeanShiftClustering

      public AlgorithmMeanShiftClustering(ModelImage image, int K, int L, int k_neigh, String data_file_name, String input_directory, int choosePoints, int jump, double percent, boolean fixedWidth, float width, boolean findOptimalKL, float epsilon, int Kmin, int Kjump, boolean FAMS_DO_SPEEDUP, int nPoints, int nDims, float[] pttemp, ModelImage modesImage, ModelImage prunedModesImage)
      File formats: Files are all ASCII files. The input file has a one line header with two number_of_points dimension Then the input points are given as lines including d numbers. The pilot process produces the neighborhood size for each point If the input file is for example d040_1S.txt and the number of neighbors is k the pilot file will be pilot_d040_1S_k.txt If the file exists the program will assume that it was created by a previous run and use the information in it and will not rerun the pilot creation procedure. This enables the users to provide their own neighborhood sizes. The output files are for example: The result of MS on the selected points is in out_data_file_name.txt The result of joined modes is in modes_data_file_name.txt Example: We are given a file d040_1S.txt fams 30 46 200 d040_1S ./ -f 0.05 10 2 Finds K,L and optionally runs mean shift on all points. To run mean shift on same data with K=24 and L=35 fams 24 35 200 d040_1S ./ NOTES: - Internally the program scales the data between 0 and 2^16-1, therefore the precision is limited by range_of_data/2^16 - Running mean shift on all the points is not necessary when only the modes are needed - In the file fams.h there are several constants that influence the speed and accuracy of the program. For example you can set the bandwidth on which two modes are joined or the number of trials on which the test is run when finding K and L. Permission to port granted by Ilan Shimshoni: No problem. Good luck with that. I'll be happy to put your code also on my website if you like. Ilan Sent from my Cyanogen phone On Nov 23, 2016 9:02 PM, "Gandler, William (NIH/CIT) [E]" invalid input: '<'ilb@mail.nih.gov> wrote: Dear Ilan Shimshoni, I am a Java programmer for the MIPAV imaging package at NIH working on mean shift. Could I have permission to port your Adaptive mean shift based clustering C++ code to Java for use in MIPAV? You will receive full acknowledgement for all code ported. Sincerely, William Gandler

  • Method Details