class CAAMOptimize

General purpose optimization of the AAM.

Inheritance:

CAAMOptimize - CAAMObject

Public Methods

inline double EvalFunction(CDVector& vX)

Fuction to be optimized.

void OptResults( CDVector &c, CAAMShape &s, double &fit )

Returns the optimisation results as c, shape and error.

Private Fields

double m_dMinFit

The minimum fit

int m_iSimilarityMeasure

The similarity measure to be used

const CDMultiBand<TAAMPixel> * m_pImage

The image where the optimization shuold be done

const CAAMModel* m_pModel

The AAM

const CAAMShape* m_pShape

The initial shape pose

CAAMShape m_sMinShape

The minimum fit shape

CDVector m_vDiff

The difference vector

CDVector m_vMinC

The minimum fit model parameters

Inherited from CAAMObject:

Public Methods

void Dump( const char* szPath ) const

void FromFile( const char* szFilename )

void FromFile( FILE* fh )

void ToFile( FILE* fh ) const

void ToFile( const char* szFilename ) const

Documentation

Class for optimizing the AAM using traditional optimization methods such as simluated annealingm, conjugated gradient, steepest descent, BGFS and pattern search.

const CAAMModel* m_pModel

The AAM

const CAAMShape* m_pShape

The initial shape pose

const CDMultiBand<TAAMPixel> * m_pImage

The image where the optimization shuold be done

int m_iSimilarityMeasure

The similarity measure to be used

CDVector m_vDiff

The difference vector

double m_dMinFit

The minimum fit

CAAMShape m_sMinShape

The minimum fit shape

CDVector m_vMinC

The minimum fit model parameters

inline double EvalFunction(CDVector& vX)

Function providing a scalar interpretation of the AAM fit based on a set of paramters.

Parameters:

vX - The independent parameters that will be optimized. The first n-4 elements constitues the normal model parameters of the AAM. The last four are the pose parameters.

Returns:

The scalar fit of the AAM.

Author:

Mikkel B. Stegmann

Version:

7-7-2000

See Also:

void OptResults( CDVector &c, CAAMShape &s, double &fit )

Returns the optimisation results as c, shape and error. This is to avoid and extra conversion after ended optimisation, and worse, an extra image sampling to get the error.

Parameters:

c - The optimal model parameters.
s - The optimal shape.
s - The optimal error.

Returns:

Nothing.

Author:

Mikkel B. Stegmann

Version:

7-7-2000

This class has no child classes.

Author:

Mikkel B. Stegmann

Version:

7-8-2000

Alphabetic index Hierarchy of classes