MatrixOps.H File Reference

---

Detailed Description

Matrix operations on Image

Definition in file MatrixOps.H.

#include "Image/Image.H"
#include "Util/Promotions.H"
#include "rutz/error.h"

Include dependency graph for MatrixOps.H:

This graph shows which files directly or indirectly include this file:

Go to the source code of this file.

Classes
class	SingularMatrixException
	Exception class thrown if a singular matrix is encountered. More...
Functions
template<class T>
Image< typename promote_trait< T, T >::TP >	vmMult (const Image< T > &v, const Image< T > &M)
	Vector-Matrix Multiplication: y = v * M.
template<class T>
Image< typename promote_trait< T, T >::TP >	matrixMult (const Image< T > &A, const Image< T > &B)
	Matrix-Matrix Multiplication: C = A * B.
template<class T>
Image< typename promote_trait< T, T >::TP >	matrixMult (const Image< T > &A, const Image< T > &B, const uint beginAX, const uint endAX, const uint beginBX, const uint endBX, const uint beginAY, const uint endAY)
	Matrix-Matrix Multiplication: C = A * B.
template<class T_or_RGB>
Image< T_or_RGB >	transpose (const Image< T_or_RGB > &M)
	transpose matrix M
template<class T_or_RGB>
Image< T_or_RGB >	flipHoriz (const Image< T_or_RGB > &img)
	flip horizontally
template<class T_or_RGB>
Image< T_or_RGB >	flipVertic (const Image< T_or_RGB > &img)
	flip vertically
template<class T>
Image< T >	eye (const uint size)
	return the identity matrix of dimensions (size x size)
template<class T>
promote_trait< T, T >::TP	trace (const Image< T > &M)
	Compute the trace of a square matrix, i.e., sum of its diagonal elements.
template<class T>
int	matrixPivot (Image< T > &M, const int y)
	Get a partial pivot and pivote a square matrix, at y.
template<class T>
Image< typename promote_trait< T, float >::TP >	matrixInv (const Image< T > &M)
	Inverse a nonsingular square matrix.
template<class T>
promote_trait< T, T >::TP	dotprod (const Image< T > &A, const Image< T > &B)
	Compute dot product between two images.
template<class T>
promote_trait< T, float >::TP	matrixDet (const Image< T > &M)
	Compute the determinant of a square matrix.

---

Function Documentation

template<class T> promote_trait<T,T>::TP dotprod (  const Image< T > &  A, const Image< T > &  B )

Compute dot product between two images. The two images must have same dims. This is just the sum of all pointwise products. Definition at line 453 of file MatrixOps.C. References ASSERT, Image< T >::begin(), Image< T >::end(), GVX_TRACE, Image< T >::isSameSize(), and lobot::stop(). Referenced by main().

template<class T> Image<T> eye (  const uint  size  )

return the identity matrix of dimensions (size x size) Definition at line 301 of file MatrixOps.C. References Image< T >::beginw(), GVX_TRACE, i, and ZEROS. Referenced by SwpeScorer::accum(), AffineTransform::computeTransform(), SimulationViewerCompress::getTraj(), WinnerTakeAllGreedy::integrate(), main(), SimulationViewerEyeSim::onSimEventWTAwinner(), and submain().

template<class T_or_RGB> Image<T_or_RGB> flipHoriz (  const Image< T_or_RGB > &  img  )

flip horizontally Definition at line 259 of file MatrixOps.C. References Image< T >::begin(), Image< T >::beginw(), Image< T >::getHeight(), Image< T >::getWidth(), GVX_TRACE, h, NO_INIT, w, x, and y. Referenced by GSlocalizer::getBeliefImage(), Beobot2_GistSalLocalizerMasterI::getBeliefImage(), RetinaStd::getOutput(), RetinaSpaceVariant::getOutput(), and replicateHemifield().

template<class T_or_RGB> Image<T_or_RGB> flipVertic (  const Image< T_or_RGB > &  img  )

flip vertically Definition at line 279 of file MatrixOps.C. References Image< T >::begin(), Image< T >::beginw(), Image< T >::getHeight(), Image< T >::getWidth(), GVX_TRACE, h, NO_INIT, w, and y. Referenced by Geons3D::calcGeonLikelihood(), TwoHalfDSketch::calcSketchLikelihood(), HippocampusI::displayMap(), BeeStemSim::getImage(), RetinaStd::getOutput(), RetinaSpaceVariant::getOutput(), linePlot(), main(), replicateHemifield(), IRobotI::run(), BeoBotSim::simLoop(), SubSim::simLoop(), IRobotSim::simLoop(), SeaBee3Simulator::simLoop(), and SeaBee3Simulator::updateSensors().

template<class T> promote_trait<T,float>::TP matrixDet (  const Image< T > &  M  )

Compute the determinant of a square matrix. Definition at line 468 of file MatrixOps.C. References ASSERT, Image< T >::getVal(), Image< T >::getWidth(), GVX_TRACE, i, Image< T >::isSquare(), matrixPivot(), and Image< T >::setVal(). Referenced by Image_xx_matrixDet_xx_1(), Image_xx_matrixDet_xx_2(), and Image_xx_matrixDet_xx_3().

template<class T> Image<typename promote_trait<T, float>::TP> matrixInv (  const Image< T > &  M  )

Inverse a nonsingular square matrix. Definition at line 373 of file MatrixOps.C. References ASSERT, Image< T >::getArrayPtr(), Image< T >::getWidth(), GVX_TRACE, i, Image< T >::isSquare(), matrixPivot(), and SRC_POS. Referenced by TaskRelevanceMapGistClassify::computeGistDist(), getLikelyhoodImage(), Image_xx_matrixInv_xx_1(), Image_xx_matrixInv_xx_2(), Image_xx_matrixInv_xx_3(), naiveUnstablePseudoInv(), and HippocampusI::updateParticleSlamObservation().

template<class T> Image<typename promote_trait<T,T>::TP> matrixMult (  const Image< T > &  A, const Image< T > &  B, const uint  beginAX, const uint  endAX, const uint  beginBX, const uint  endBX, const uint  beginAY, const uint  endAY )

Matrix-Matrix Multiplication: C = A * B. The images A and B are interpreted as matrices. For multiplying a (m x n) with a (n x p) matrix, this is the generic matrix multiplication with complexity (m * n * p). This version will allow you to multiply values in a matrix that lie between index endA/B and beginA/B. The destination matrix will be the same size as: (endAX - beginAX) x (endAX - beginAX) per usual the width of A (indexed between beginAX and endAX) and height of B MUST be equal. If not then perhaps use transpose... Parameters: A  input matrix 1 B  input matrix 2 beginAX  starting (width) index in matrix A endAX  ending (width) index in matrix A beginBX  starting (width) index in matrix B endBX  ending (width) index in matrix B beginAY  starting (height) index in matrix A endAY  ending (height) index in matrix A Definition at line 207 of file MatrixOps.C. References ASSERT, Image< T >::beginw(), Image< T >::getHeight(), Image< T >::getVal(), GVX_TRACE, Image< T >::initialized(), and ZEROS.

template<class T> Image<typename promote_trait<T,T>::TP> matrixMult (  const Image< T > &  A, const Image< T > &  B )

Matrix-Matrix Multiplication: C = A * B. The images A and B are interpreted as matrices. For multiplying a (m x n) with a (n x p) matrix, this is the generic matrix multiplication with complexity (m * n * p). Definition at line 179 of file MatrixOps.C. References basic_mm_mul().

template<class T> int matrixPivot (  Image< T > &  M, const int  y )

Get a partial pivot and pivote a square matrix, at y. returns -1 if we can't find one Definition at line 338 of file MatrixOps.C. References abs(), ASSERT, Image< T >::getArrayPtr(), Image< T >::getVal(), Image< T >::getWidth(), GVX_TRACE, i, and Image< T >::isSquare(). Referenced by matrixDet(), and matrixInv().

template<class T> promote_trait<T,T>::TP trace (  const Image< T > &  M  )

Compute the trace of a square matrix, i.e., sum of its diagonal elements. Definition at line 319 of file MatrixOps.C. References ASSERT, Image< T >::begin(), Image< T >::getWidth(), GVX_TRACE, i, and Image< T >::isSquare(). Referenced by invdiag().

template<class T_or_RGB> Image<T_or_RGB> transpose (  const Image< T_or_RGB > &  M  )

transpose matrix M Definition at line 235 of file MatrixOps.C. References Image< T >::begin(), Image< T >::beginw(), Image< T >::getHeight(), Image< T >::getWidth(), GVX_TRACE, h, NO_INIT, w, x, and y. Referenced by FeedForwardNetwork::backprop(), FeedForwardNetwork::backprop3L(), TaskRelevanceMapGistClassify::computeGistDist(), GSlocalizer::getBeliefImage(), Beobot2_GistSalLocalizerMasterI::getBeliefImage(), BackpropLearner::getBiasMap(), TaskRelevanceMapTigs2::getGistPCAMatrix(), TaskRelevanceMapTigs2::getImgPCAMatrix(), getLikelyhoodImage(), TaskRelevanceMapGistClassify::getPCAMatrix(), TaskRelevanceMapTigs::getPCAMatrix(), TaskRelevanceMapTigs::getTigsMatrix(), TaskRelevanceMapTigs2::getTigsMatrix(), Image_xx_svd_lapack_xx_2(), ImageInfo::ImageInfo(), FeedForwardNetwork::init(), KalmanFilter::init(), FeedForwardNetwork::init3L(), TaskRelevanceMapTigs2::inputFrame(), linePlot(), main(), naiveUnstablePseudoInv(), lapack::svd(), lapack::svdf(), svdIdent(), svdPseudoInv(), svdPseudoInvf(), BackpropNetwork::train(), HippocampusI::updateParticleSlamObservation(), FeedForwardNetwork::write(), and FeedForwardNetwork::write3L().

template<class T> Image<typename promote_trait<T,T>::TP> vmMult (  const Image< T > &  v, const Image< T > &  M )

Vector-Matrix Multiplication: y = v * M. Parameters: v  row vector (w x h == n x 1), M  matrix with (w x h == p x n) Returns: row vector (w x h == p x 1) Definition at line 151 of file MatrixOps.C. References basic_vm_mul().

---