Main Page   Modules   Compound List   File List   Compound Members   File Members  

fast_hough_transform.h File Reference

#include <gandalf/common/bit_array.h>
#include <gandalf/linalg/mat_gen.h>

Go to the source code of this file.

Functions

Gan_Bool gan_fast_hough_transform (int k, Gan_Matrix *a, int *weight, int no_points, double *S0, double *X0, int max_level, int T_thres, Gan_MemoryStack *memory_stack, double *X_best, int *level_best, int *accum_best, Gan_BitArray *list_best, int *subdivs)
 General purpose Fast Hough Transform (FHT) function.


Detailed Description

Module: Fast Hough Transform implentation

Part of: Gandalf Library

Revision: Last edited: Author:

Copyright: (c) 2000 Imagineer Software Limited


Function Documentation

Gan_Bool gan_fast_hough_transform int    k,
Gan_Matrix   a,
int *    weight,
int    no_points,
double *    S0,
double *    X0,
int    max_level,
int    T_thres,
Gan_MemoryStack   memory_stack,
double *    X_best,
int *    level_best,
int *    accum_best,
Gan_BitArray   list_best,
int *    subdivs
 

General purpose Fast Hough Transform (FHT) function.

Parameters:
k Dimensionality of parameter space
a Feature space (k+1)-vectors arranged as a matrix
weight Weights assigned to each feature vector
no_points Number of data points
S0 Half-range of parameters
X0 Centre of parameter ranges
max_level Maximum subdivision level
T_thres Threshold T in FHT for deciding whether subdivide or not
memory_stack Pointer to memory stack structure
X_best Output best-fit parameters
level_best Output subdivision level reached by FHT
accum_best Output sum of weights of points with lines intersecting final hypersphere
list_best Output bit array bit-list, with 1's for points involved in best-fit result
subdivs Output total number of subdivisions
Returns:
GAN_TRUE on success, GAN_FALSE on failure.
For each feature point j the relationship between feature space uses parametrisation

(not normalised) where X[i] are the parameters and k is the dimensionality of parameter space. The starting parameter half-ranges are given by the S0 vector, and the parameter origin by the X0 vector. S0 and X0 thus define the root hypercube:

This is a depth-first version of the FHT, i.e. the child hypercubes are subdivided exhaustively before trying another child. This minimises memory requirement by limiting it to

a contains feature space (k+1)-vectors arranged as a no_points by k+1 matrix.


Generated on Mon Oct 13 16:14:41 2003 by doxygen1.3-rc1