Illumination Model is a parametric function that transforms pixel values extracted from a patch to account for illumination changes g(I, a): R(N) x R(K) -> R(N) More...

#include <IlluminationModel.h>

Inheritance diagram for IlluminationModel:

Public Types
enum	PixHessType { Constant, Diagonal, General }

Public Member Functions
	IlluminationModel (const ILMParams *ilm_params)

virtual int	getStateSize () const =0

virtual void	initialize (double *p)

virtual void	apply (double g, const double I, const double *p)=0

virtual void	invert (double inv_p, const double p)=0

virtual void	update (double new_p, const double old_p, const double *dp)=0

virtual void	cmptParamJacobian (double df_dp, const double df_dg, const double I, const double p)=0
	ILM Jacobians.

virtual void	cmptPixJacobian (double df_dI, const double df_dg, const double I, const double p)=0

virtual void	cmptParamHessian (double d2f_dp2, const double d2f_dg2, const double df_dg, const double I, const double *p)=0
	NULL/nullptr for d2f_dg2 implies that the corresponding matrix is identity; conversely a non negative return imples an identity d2f_dI2 scaled by the return value; these conventions are necessary to avoid unnecessary and very expensive matrix multiplications;. More...

virtual void	cmptPixHessian (double d2f_dI2, const double d2f_dg2, const double df_dg, const double I, const double *p)=0
	d2f_dI2 = dg_dI^Td2f_dg2dg_dI + df_dg*d2g_dI2

virtual void	cmptCrossHessian (double d2f_dp_dI, const double d2f_dg2, const double df_dg, const double I, const double *p)=0
	d2f_dp_dI = dg_dp^Td2f_dg2dg_dI + df_dg*d2g_dpdI

virtual void	cmptParamHessian (double d2f_dp2, const double d2f_dg2, const double I, const double p)=0
	First order ILM hessians. More...

virtual void	cmptPixHessian (double d2f_dI2, const double d2f_dg2, const double I, const double p)=0
	d2f_dI2 = dg_dI^Td2f_dg2dg_dI

virtual void	cmptCrossHessian (double d2f_dp_dI, const double d2f_dg2, const double I, const double p)=0
	d2f_dp_dI = dg_dp^Td2f_dg2dg_dI

virtual void	setPixHessType (PixHessType _d2f_dg2_type)

virtual PixHessType	getPixHessType ()=0

virtual void	parseSamplerSigma (VectorXd &out_sigma, const VectorXd &in_sigma)=0

virtual void	parseSamplerMean (VectorXd &out_mean, const VectorXd &in_mean)=0

Public Attributes
string	name

bool	apply_on_init

Protected Attributes
int	resx

int	resy

int	n_pix

PixHessType	d2f_dg2_type

Detailed Description

Illumination Model is a parametric function that transforms pixel values extracted from a patch to account for illumination changes g(I, a): R(N) x R(K) -> R(N)

Member Function Documentation

virtual void IlluminationModel::cmptParamHessian	(	double *	d2f_dp2,
		const double *	d2f_dg2,
		const double *	df_dg,
		const double *	I,
		const double *	p
	)

pure virtual

NULL/nullptr for d2f_dg2 implies that the corresponding matrix is identity; conversely a non negative return imples an identity d2f_dI2 scaled by the return value; these conventions are necessary to avoid unnecessary and very expensive matrix multiplications;.

Second order ILM hessiansd2f_dp2 = dg_dp^T*d2f_dg2*dg_dp + df_dg*d2g_dp2

Implemented in PGB, RBF, and GB.

virtual void IlluminationModel::cmptParamHessian	(	double *	d2f_dp2,
		const double *	d2f_dg2,
		const double *	I,
		const double *	p
	)

pure virtual

First order ILM hessians.

d2f_dp2 = dg_dp^T*d2f_dg2*dg_dp

Implemented in PGB, RBF, and GB.

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

AM/include/mtf/AM/IlluminationModel.h

Public Types

Public Member Functions

Public Attributes

Protected Attributes

Detailed Description

Member Function Documentation