Public Member Functions | Static Public Member Functions | Protected Member Functions | Protected Attributes
rgrl_transformation Class Reference

A base class that represents a transformation. More...

#include <rgrl_transformation.h>

Inheritance diagram for rgrl_transformation:
Inheritance graph
[legend]

List of all members.

Public Member Functions

virtual ~rgrl_transformation ()
 rgrl_transformation ()
 default constructor.
 rgrl_transformation (const vnl_matrix< double > &cov)
 initialize with covariance matrix.
void map_location (vnl_vector< double > const &from, vnl_vector< double > &to) const
 Apply the transformation to create a new (mapped) location.
vnl_vector< double > map_location (vnl_vector< double > const &p) const
 Apply the transformation to create a new (mapped) location.
virtual void map_tangent (vnl_vector< double > const &from_loc, vnl_vector< double > const &from_dir, vnl_vector< double > &to_dir) const
 Map a tangent direction.
virtual void map_normal (vnl_vector< double > const &from_loc, vnl_vector< double > const &from_dir, vnl_vector< double > &to_dir) const
 Map a normal direction.
virtual void map_normal (vnl_vector< double > const &from_loc, vnl_vector< double > const &from_dir, vnl_matrix< double > const &tangent_subspace, vnl_vector< double > &to_dir) const
 Map a normal direction, given the tangent subspace.
void map_direction (vnl_vector< double > const &from_loc, vnl_vector< double > const &from_dir, vnl_vector< double > &to_dir) const
 Map an arbitrary direction which is neither a tangent nor a normal.
virtual double map_intensity (vnl_vector< double > const &from, double intensity) const
 Apply to an intensity, with a default of the identity.
virtual vnl_matrix< double > transfer_error_covar (vnl_vector< double > const &p) const =0
 Compute covariance of the transfer error based on transformation covariance.
virtual void inv_map (const vnl_vector< double > &to, bool initialize_next, const vnl_vector< double > &to_delta, vnl_vector< double > &from, vnl_vector< double > &from_next_est) const
 Inverse map with an initial guess.
vnl_matrix< double > covar () const
 Parameter covariance matrix.
virtual double log_det_covar () const
 log of determinant of the covariance.
void set_covar (const vnl_matrix< double > &covar)
 set parameter covariance matrix.
bool is_covar_set () const
 is covariance set?.
virtual void inv_map (const vnl_vector< double > &to, vnl_vector< double > &from) const
 Inverse map based on the transformation.
virtual bool is_invertible () const
 is this an invertible transformation?.
virtual rgrl_transformation_sptr inverse_transform () const
 Return an inverse transformation.
vnl_matrix< double > jacobian (vnl_vector< double > const &from_loc) const
 Return the jacobian of the transform.
virtual void jacobian_wrt_loc (vnl_matrix< double > &jac, vnl_vector< double > const &from_loc) const =0
 Compute jacobian w.r.t. location.
virtual rgrl_transformation_sptr scale_by (double scale) const =0
 transform the transformation for images of different resolution.
virtual void write (vcl_ostream &os) const
 output transformation.
virtual bool read (vcl_istream &is)
 input transformation.
virtual rgrl_transformation_sptr clone () const =0
 make a clone copy.
void set_scaling_factors (vnl_vector< double > const &scaling)
 set scaling factors.
const vnl_vector< double > & scaling_factors () const
 return scaling factor.
 rgrl_type_macro (rgrl_transformation, rgrl_object)
virtual bool is_type (const vcl_type_info &type) const
void set_debug_flag (unsigned int debugFlag) const
 Set the value of the debug flag. A non-zero value turns debugging on.
unsigned int debug_flag () const
 Get the value of the debug flag.
void set_warning (bool) const
 Set the flag for warning messages.
bool warning () const
 Get the warning flag.
unsigned int add_observer (rgrl_event_sptr event, rgrl_command_sptr)
 Allow people to add/remove/invoke observers (callbacks) to any rgrl object.
rgrl_command_sptr get_command (unsigned int tag)
 Get the command associated with the given tag.
void invoke_event (const rgrl_event &)
 Call execute(.) on all the rgrl_commands observing this event id.
void invoke_event (const rgrl_event &) const
 Call execute(.) on all the rgrl_commands observing this event id.
void remove_observer (unsigned int tag)
 Remove the observer with this tag value.
bool has_observer (const rgrl_event &event) const
 Return true if an observer is registered for this event.
void ref ()
void unref ()
int get_references () const
bool is_referenced () const

Static Public Member Functions

static const vcl_type_info & type_id ()

Protected Member Functions

virtual void map_loc (vnl_vector< double > const &from, vnl_vector< double > &to) const =0
 Apply the transformation to create a new (mapped) location.
virtual void map_dir (vnl_vector< double > const &from_loc, vnl_vector< double > const &from_dir, vnl_vector< double > &to_dir) const =0
 Apply the transformation to create a new direction at the (mapped) location.
double log_det_sym_matrix (vnl_matrix< double > const &m) const
 a slightly better way of computing log of determinant.
double log_det_covar_deficient (int rank) const
 compute the log of determinant of the covariance when the matrix is rank deficient.

Protected Attributes

vnl_matrix< double > covar_
 covariance matrix.
bool is_covar_set_
 flag of setting covariance.
vnl_vector< double > scaling_factors_
 scaling factors of current transformation on each dimension.

Detailed Description

A base class that represents a transformation.

Derived classes must implement

We use different names for the functions that derived classes implement because otherwise the "name hiding rule" of the C++ standard is invoked, which means the some of the functionality in the base class interface--in particular, the overloaded convenience functions--are unavailable in the derived class interface.

Definition at line 33 of file rgrl_transformation.h.


Constructor & Destructor Documentation

rgrl_transformation::~rgrl_transformation ( ) [virtual]

Definition at line 29 of file rgrl_transformation.cxx.

rgrl_transformation::rgrl_transformation ( ) [inline]

default constructor.

Definition at line 40 of file rgrl_transformation.h.

rgrl_transformation::rgrl_transformation ( const vnl_matrix< double > &  cov)

initialize with covariance matrix.

Definition at line 22 of file rgrl_transformation.cxx.


Member Function Documentation

unsigned int rgrl_object::add_observer ( rgrl_event_sptr  event,
rgrl_command_sptr  cmd 
) [inherited]

Allow people to add/remove/invoke observers (callbacks) to any rgrl object.

This is an implementation of the subject/observer design pattern. An observer is added by specifying an event to respond to and an rgrl_command to execute. It returns an unsigned long tag which can be used later to remove the event or retrieve the command.

Definition at line 47 of file rgrl_object.cxx.

virtual rgrl_transformation_sptr rgrl_transformation::clone ( ) const [pure virtual]
vnl_matrix< double > rgrl_transformation::covar ( ) const

Parameter covariance matrix.

Definition at line 325 of file rgrl_transformation.cxx.

unsigned int rgrl_object::debug_flag ( ) const [inherited]

Get the value of the debug flag.

Definition at line 26 of file rgrl_object.cxx.

rgrl_command_sptr rgrl_object::get_command ( unsigned int  tag) [inherited]

Get the command associated with the given tag.

Definition at line 57 of file rgrl_object.cxx.

bool rgrl_object::has_observer ( const rgrl_event event) const [inherited]

Return true if an observer is registered for this event.

Definition at line 107 of file rgrl_object.cxx.

void rgrl_transformation::inv_map ( const vnl_vector< double > &  to,
bool  initialize_next,
const vnl_vector< double > &  to_delta,
vnl_vector< double > &  from,
vnl_vector< double > &  from_next_est 
) const [virtual]
void rgrl_transformation::inv_map ( const vnl_vector< double > &  to,
vnl_vector< double > &  from 
) const [virtual]

Inverse map based on the transformation.

This function only exist for certain transformations.

Reimplemented in rgrl_trans_quadratic, rgrl_trans_reduced_quad, rgrl_trans_spline, rgrl_trans_affine, rgrl_trans_rigid, rgrl_trans_similarity, rgrl_trans_rad_dis_homo2d, rgrl_trans_couple, rgrl_trans_translation, and rgrl_trans_homography2d.

Definition at line 598 of file rgrl_transformation.cxx.

rgrl_transformation_sptr rgrl_transformation::inverse_transform ( ) const [virtual]

Return an inverse transformation.

This function only exist for certain transformations.

Reimplemented in rgrl_trans_quadratic, rgrl_trans_reduced_quad, rgrl_trans_spline, rgrl_trans_affine, rgrl_trans_rigid, rgrl_trans_similarity, rgrl_trans_couple, rgrl_trans_rad_dis_homo2d, rgrl_trans_translation, and rgrl_trans_homography2d.

Definition at line 610 of file rgrl_transformation.cxx.

void rgrl_object::invoke_event ( const rgrl_event event) [inherited]

Call execute(.) on all the rgrl_commands observing this event id.

Definition at line 69 of file rgrl_object.cxx.

void rgrl_object::invoke_event ( const rgrl_event event) const [inherited]

Call execute(.) on all the rgrl_commands observing this event id.

The actions triggered by this call doesn't modify this object.

Definition at line 83 of file rgrl_object.cxx.

bool rgrl_transformation::is_covar_set ( ) const [inline]

is covariance set?.

Definition at line 131 of file rgrl_transformation.h.

virtual bool rgrl_transformation::is_invertible ( ) const [inline, virtual]
virtual bool rgrl_object::is_type ( const vcl_type_info &  type) const [inline, virtual, inherited]

Reimplemented in rgrl_scale_estimator.

Definition at line 66 of file rgrl_object.h.

vnl_matrix< double > rgrl_transformation::jacobian ( vnl_vector< double > const &  from_loc) const

Return the jacobian of the transform.

use only as backward compatibility

Definition at line 333 of file rgrl_transformation.cxx.

virtual void rgrl_transformation::jacobian_wrt_loc ( vnl_matrix< double > &  jac,
vnl_vector< double > const &  from_loc 
) const [pure virtual]
double rgrl_transformation::log_det_covar ( ) const [virtual]

log of determinant of the covariance.

Reimplemented in rgrl_trans_rad_dis_homo2d, rgrl_trans_homography2d, and rgrl_trans_homo2d_proj_rad.

Definition at line 231 of file rgrl_transformation.cxx.

double rgrl_transformation::log_det_covar_deficient ( int  rank) const [protected]

compute the log of determinant of the covariance when the matrix is rank deficient.

Definition at line 258 of file rgrl_transformation.cxx.

double rgrl_transformation::log_det_sym_matrix ( vnl_matrix< double > const &  m) const [protected]

a slightly better way of computing log of determinant.

Definition at line 238 of file rgrl_transformation.cxx.

virtual void rgrl_transformation::map_dir ( vnl_vector< double > const &  from_loc,
vnl_vector< double > const &  from_dir,
vnl_vector< double > &  to_dir 
) const [protected, pure virtual]

Apply the transformation to create a new direction at the (mapped) location.

The resulting direction to_dir is a unit vector.

Implemented in rgrl_trans_affine, rgrl_trans_quadratic, rgrl_trans_rad_dis_homo2d, rgrl_trans_rigid, rgrl_trans_similarity, rgrl_trans_couple, rgrl_trans_reduced_quad, rgrl_trans_homo2d_proj_rad, rgrl_trans_homography2d, rgrl_trans_translation, rgrl_trans_spline, and rgrl_trans_mixed_spline.

void rgrl_transformation::map_direction ( vnl_vector< double > const &  from_loc,
vnl_vector< double > const &  from_dir,
vnl_vector< double > &  to_dir 
) const

Map an arbitrary direction which is neither a tangent nor a normal.

The resulting direction to_dir is a unit vector.

Definition at line 52 of file rgrl_transformation.cxx.

double rgrl_transformation::map_intensity ( vnl_vector< double > const &  from,
double  intensity 
) const [virtual]

Apply to an intensity, with a default of the identity.

Definition at line 223 of file rgrl_transformation.cxx.

virtual void rgrl_transformation::map_loc ( vnl_vector< double > const &  from,
vnl_vector< double > &  to 
) const [protected, pure virtual]
void rgrl_transformation::map_location ( vnl_vector< double > const &  from,
vnl_vector< double > &  to 
) const

Apply the transformation to create a new (mapped) location.

Definition at line 35 of file rgrl_transformation.cxx.

vnl_vector< double > rgrl_transformation::map_location ( vnl_vector< double > const &  p) const

Apply the transformation to create a new (mapped) location.

Convenience call to the two parameter version that allocates the result vector.

Definition at line 43 of file rgrl_transformation.cxx.

void rgrl_transformation::map_normal ( vnl_vector< double > const &  from_loc,
vnl_vector< double > const &  from_dir,
vnl_vector< double > &  to_dir 
) const [virtual]

Map a normal direction.

The resulting direction to_dir is a unit vector.

Reimplemented in rgrl_trans_couple.

Definition at line 76 of file rgrl_transformation.cxx.

void rgrl_transformation::map_normal ( vnl_vector< double > const &  from_loc,
vnl_vector< double > const &  from_dir,
vnl_matrix< double > const &  tangent_subspace,
vnl_vector< double > &  to_dir 
) const [virtual]

Map a normal direction, given the tangent subspace.

The resulting direction to_dir is a unit vector.

Reimplemented in rgrl_trans_couple.

Definition at line 171 of file rgrl_transformation.cxx.

void rgrl_transformation::map_tangent ( vnl_vector< double > const &  from_loc,
vnl_vector< double > const &  from_dir,
vnl_vector< double > &  to_dir 
) const [virtual]

Map a tangent direction.

The resulting direction to_dir is a unit vector.

Reimplemented in rgrl_trans_couple.

Definition at line 61 of file rgrl_transformation.cxx.

bool rgrl_transformation::read ( vcl_istream &  is) [virtual]
void rgrl_object::remove_observer ( unsigned int  tag) [inherited]

Remove the observer with this tag value.

Definition at line 97 of file rgrl_object.cxx.

rgrl_transformation::rgrl_type_macro ( rgrl_transformation  ,
rgrl_object   
)
virtual rgrl_transformation_sptr rgrl_transformation::scale_by ( double  scale) const [pure virtual]
const vnl_vector<double>& rgrl_transformation::scaling_factors ( ) const [inline]

return scaling factor.

Definition at line 170 of file rgrl_transformation.h.

void rgrl_transformation::set_covar ( const vnl_matrix< double > &  covar)

set parameter covariance matrix.

Reimplemented in rgrl_trans_spline.

Definition at line 344 of file rgrl_transformation.cxx.

void rgrl_object::set_debug_flag ( unsigned int  debugFlag) const [inherited]

Set the value of the debug flag. A non-zero value turns debugging on.

Definition at line 19 of file rgrl_object.cxx.

void rgrl_transformation::set_scaling_factors ( vnl_vector< double > const &  scaling)

set scaling factors.

Unless the transformation is not estimated using estimators in rgrl, it does not need to be set explicitly

Definition at line 357 of file rgrl_transformation.cxx.

void rgrl_object::set_warning ( bool  flag) const [inherited]

Set the flag for warning messages.

Definition at line 33 of file rgrl_object.cxx.

virtual vnl_matrix<double> rgrl_transformation::transfer_error_covar ( vnl_vector< double > const &  p) const [pure virtual]

Compute covariance of the transfer error based on transformation covariance.

This gives the additional uncertainty of the transferred point location due to the uncertainty of the transform estimate.

Implemented in rgrl_trans_quadratic, rgrl_trans_rigid, rgrl_trans_reduced_quad, rgrl_trans_affine, rgrl_trans_homo2d_proj_rad, rgrl_trans_couple, rgrl_trans_rad_dis_homo2d, rgrl_trans_similarity, rgrl_trans_translation, rgrl_trans_homography2d, rgrl_trans_spline, and rgrl_trans_mixed_spline.

static const vcl_type_info& rgrl_object::type_id ( ) [inline, static, inherited]

Reimplemented in rgrl_scale_estimator.

Definition at line 63 of file rgrl_object.h.

bool rgrl_object::warning ( ) const [inherited]

Get the warning flag.

Definition at line 40 of file rgrl_object.cxx.

void rgrl_transformation::write ( vcl_ostream &  os) const [virtual]

Member Data Documentation

covariance matrix.

Unlike transformation parameters, covariance is always a mtrix of double.

Definition at line 205 of file rgrl_transformation.h.

flag of setting covariance.

Check it before using covariance matrix

Definition at line 209 of file rgrl_transformation.h.

scaling factors of current transformation on each dimension.

This is computed from current transformation. And it has nothing to do with how to transform points

Definition at line 214 of file rgrl_transformation.h.


The documentation for this class was generated from the following files: