TemplateTrackerMIInverseCompositional¶

class TemplateTrackerMIInverseCompositional(self)¶

Bases: TemplateTrackerMI

The algorithm implemented in this class is described in [12] and [31] .

Default constructor.

Methods

__init__	Default constructor.
initTemplateRefBspline	return: A tuple containing:
setMinimizationMethod
setUseTemplateSelect	Use only the strong gradient pixels to compute the Jabobian.

Inherited Methods

initClick	Select the reference template in image I using mouse click.
setCostFunctionVerification
setHDes
USE_HESSIEN_BEST_COND
HESSIAN_0
BsplineType	Hessian computation.
setGain
BSPLINE_THIRD_ORDER
getRatioPixelIn
setThresholdGradient
setNbIterBrent
setSampling	Set the pixel sampling parameters along the rows and the columns.
setNc
getNbIteration
setHessianComputation
initFromZone	Initialize the reference template from a vector of points.
trackRobust
setCovarianceComputation
BSPLINE_FOURTH_ORDER
setBspline
HessienType	Hessian computation.
getCovarianceMatrix
resetTracker	Reset the tracker by freeing the memory allocated by the template tracker during the initialization.
getNMI
setBlur
USE_HESSIEN_NORMAL
setApprocHessian
HESSIAN_NONSECOND
getZoneRef	Return the reference template zone.
setGaussianFilterSize
getMI	Overloaded function.
HESSIAN_YOUCEF
initFromPoints	Initialize the reference template from a vector of points.
getNbParam
getH
getG
getdp
setUseBrent	By default Brent usage is disabled.
HESSIAN_EXACT
track	Track the template on image I .
HESSIAN_d2I
setPyramidal	Set the number of pyramid levels used in the multi-resolution scheme.
HessienApproximationType	Hessian computation.
setIterationMax	Set the maximum number of iteration of the estimation scheme.
getDiverge
display	Overloaded function.
getp
setLambda
getMI256
HESSIAN_NEW
USE_HESSIEN_DESIRE
setp
setThresholdResidualDifference	Set the threshold used to stop optimization loop.

Operators

__doc__
__init__	Default constructor.
__module__

Attributes

BSPLINE_FOURTH_ORDER
BSPLINE_THIRD_ORDER
HESSIAN_0
HESSIAN_EXACT
HESSIAN_NEW
HESSIAN_NONSECOND
HESSIAN_YOUCEF
HESSIAN_d2I
USE_GRADIENT
USE_HESSIEN_BEST_COND
USE_HESSIEN_DESIRE
USE_HESSIEN_NORMAL
USE_LMA
USE_NEWTON
USE_QUASINEWTON
__annotations__

class BsplineType(self, value: int)¶

Bases: pybind11_object

Hessian computation.

Values:

• BSPLINE_THIRD_ORDER

• BSPLINE_FOURTH_ORDER

__and__(self, other: object) → object¶

__eq__(self, other: object) → bool¶

__ge__(self, other: object) → bool¶

__getstate__(self) → int¶

__gt__(self, other: object) → bool¶

__hash__(self) → int¶

__index__(self) → int¶

__init__(self, value: int)¶

__int__(self) → int¶

__invert__(self) → object¶

__le__(self, other: object) → bool¶

__lt__(self, other: object) → bool¶

__ne__(self, other: object) → bool¶

__or__(self, other: object) → object¶

__rand__(self, other: object) → object¶

__ror__(self, other: object) → object¶

__rxor__(self, other: object) → object¶

__setstate__(self, state: int) → None¶

__xor__(self, other: object) → object¶

property name : str¶

class HessienApproximationType(self, value: int)¶

Bases: pybind11_object

Hessian computation.

Values:

• BSPLINE_THIRD_ORDER

• BSPLINE_FOURTH_ORDER

__and__(self, other: object) → object¶

__eq__(self, other: object) → bool¶

__ge__(self, other: object) → bool¶

__getstate__(self) → int¶

__gt__(self, other: object) → bool¶

__hash__(self) → int¶

__index__(self) → int¶

__init__(self, value: int)¶

__int__(self) → int¶

__invert__(self) → object¶

__le__(self, other: object) → bool¶

__lt__(self, other: object) → bool¶

__ne__(self, other: object) → bool¶

__or__(self, other: object) → object¶

__rand__(self, other: object) → object¶

__ror__(self, other: object) → object¶

__rxor__(self, other: object) → object¶

__setstate__(self, state: int) → None¶

__xor__(self, other: object) → object¶

property name : str¶

class HessienType(self, value: int)¶

Bases: pybind11_object

Hessian computation.

Values:

• BSPLINE_THIRD_ORDER

• BSPLINE_FOURTH_ORDER

__and__(self, other: object) → object¶

__eq__(self, other: object) → bool¶

__ge__(self, other: object) → bool¶

__getstate__(self) → int¶

__gt__(self, other: object) → bool¶

__hash__(self) → int¶

__index__(self) → int¶

__init__(self, value: int)¶

__int__(self) → int¶

__invert__(self) → object¶

__le__(self, other: object) → bool¶

__lt__(self, other: object) → bool¶

__ne__(self, other: object) → bool¶

__or__(self, other: object) → object¶

__rand__(self, other: object) → object¶

__ror__(self, other: object) → object¶

__rxor__(self, other: object) → object¶

__setstate__(self, state: int) → None¶

__xor__(self, other: object) → object¶

property name : str¶

class MinimizationTypeMIInverseCompositional(self, value: int)¶

Bases: pybind11_object

Minimization method.

Values:

• USE_NEWTON

• USE_LMA

• USE_GRADIENT

• USE_QUASINEWTON

__and__(self, other: object) → object¶

__eq__(self, other: object) → bool¶

__ge__(self, other: object) → bool¶

__getstate__(self) → int¶

__gt__(self, other: object) → bool¶

__hash__(self) → int¶

__index__(self) → int¶

__init__(self, value: int)¶

__int__(self) → int¶

__invert__(self) → object¶

__le__(self, other: object) → bool¶

__lt__(self, other: object) → bool¶

__ne__(self, other: object) → bool¶

__or__(self, other: object) → object¶

__rand__(self, other: object) → object¶

__ror__(self, other: object) → object¶

__rxor__(self, other: object) → object¶

__setstate__(self, state: int) → None¶

__xor__(self, other: object) → object¶

property name : str¶

__init__(self)¶

Default constructor.

display(*args, **kwargs)¶

Overloaded function.

1. display(self: visp._visp.tt.TemplateTracker, I: visp._visp.core.ImageGray, col: visp._visp.core.Color = vpColor::green, thickness: int = 3) -> None

Display the warped reference template in an image.

The following code shows how to use display capabilities:

#include <visp3/tt/vpTemplateTrackerSSDInverseCompositional.h>
#include <visp3/tt/vpTemplateTrackerWarpHomography.h>

#ifdef ENABLE_VISP_NAMESPACE
using namespace VISP_NAMESPACE_NAME;
#endif

int main()
{
  vpImage<unsigned char> I;
  vpTemplateTrackerWarpHomography warp;
  vpTemplateTrackerSSDInverseCompositional tracker(&warp);
  vpTemplateTrackerZone zoneRef, zoneWarped;

  // Display the warped zone
  tracker.display(I, vpColor::red);

  // Display the reference zone
  zoneRef = tracker.getZoneRef();
  zoneRef.display(I, vpColor::green);

  // Display the warped zone
  vpColVector p = tracker.getp();
  warp.warpZone(zoneRef, p, zoneWarped);
  zoneWarped.display(I, vpColor::blue);
}

Parameters:

I

Image in which the warped zone has to be displayed.

col

Color used to draw the triangle edges.

thickness

Thickness of the lines.

2. display(self: visp._visp.tt.TemplateTracker, I: visp._visp.core.ImageRGBa, col: visp._visp.core.Color = vpColor::green, thickness: int = 3) -> None

Display the warped reference template in an image.

The following code shows how to use display capabilities:

#include <visp3/tt/vpTemplateTrackerSSDInverseCompositional.h>
#include <visp3/tt/vpTemplateTrackerWarpHomography.h>

#ifdef ENABLE_VISP_NAMESPACE
using namespace VISP_NAMESPACE_NAME;
#endif

int main()
{
  vpImage<vpRGBa> I;
  vpTemplateTrackerWarpHomography warp;
  vpTemplateTrackerSSDInverseCompositional tracker(&warp);
  vpTemplateTrackerZone zoneRef, zoneWarped;

  // Display the warped zone
  tracker.display(I, vpColor::red);

  // Display the reference zone
  zoneRef = tracker.getZoneRef();
  zoneRef.display(I, vpColor::green);

  // Display the warped zone
  vpColVector p = tracker.getp();
  warp.warpZone(zoneRef, p, zoneWarped);
  zoneWarped.display(I, vpColor::blue);
}

Parameters:

I

Image in which the warped zone has to be displayed.

col

Color used to draw the triangle edges.

thickness

Thickness of the lines.

getCovarianceMatrix(self) → visp._visp.core.Matrix¶

getDiverge(self) → bool¶

getG(self) → visp._visp.core.ColVector¶

getH(self) → visp._visp.core.Matrix¶

getMI(*args, **kwargs)¶

Overloaded function.

1. getMI(self: visp._visp.tt_mi.TemplateTrackerMI) -> float

2. getMI(self: visp._visp.tt_mi.TemplateTrackerMI, I: visp._visp.core.ImageGray, nc: int, bspline: int, tp: visp._visp.core.ColVector) -> tuple[float, int]

Returns:

A tuple containing:

• nc

getMI256(self, I: visp._visp.core.ImageGray, tp: visp._visp.core.ColVector) → float¶

getNMI(self) → float¶

getNbIteration(self) → int¶

getNbParam(self) → int¶

getRatioPixelIn(self) → float¶

getZoneRef(self) → visp._visp.tt.TemplateTrackerZone¶

Return the reference template zone.

getdp(self) → visp._visp.core.ColVector¶

getp(self) → visp._visp.core.ColVector¶

initClick(self, I: visp._visp.core.ImageGray, delaunay: bool = false) → None¶

Select the reference template in image I using mouse click.

Parameters:

I: visp._visp.core.ImageGray¶

Image containing the reference template.

delaunay: bool = false¶

Flag used to enable Delaunay triangulation.

• If true, from the image points selected by the user, a Delaunay triangulation is performed to initialize the reference template.

  • A left click select a image point;

  • A right click select the last image point and ends the initialisation stage.

• If false, the user select directly points as successive triangle corners. The size of v_ip vector should be a multiple of 3. It is not mandatory that triangles have one edge in common; they can define a discontinued area.

  • A left click select a triangle corner;

  • A right click select the last triangle corner and ends the initialisation stage. For example, to select the reference template as two triangles, the user has to left click five times and finish the selection on the sixth corner with a right click.

initFromPoints(self, I: visp._visp.core.ImageGray, v_ip: list[visp._visp.core.ImagePoint], delaunay: bool = false) → None¶

Initialize the reference template from a vector of points.

Parameters:

I: visp._visp.core.ImageGray¶

Image containing the reference template.

v_ip: list[visp._visp.core.ImagePoint]¶

Vector of image points defining the reference template.

delaunay: bool = false¶

• If true, from the image points defining the reference template enable Delaunay triangulation.

• If false, the vector of image points define the reference template as a list of triangles. The size of v_ip vector should be a multiple of 3.

initFromZone(self, I: visp._visp.core.ImageGray, zone: visp._visp.tt.TemplateTrackerZone) → None¶

Initialize the reference template from a vector of points.

Parameters:

I: visp._visp.core.ImageGray¶

Image containing the reference template.

zone: visp._visp.tt.TemplateTrackerZone¶

The zone that describes the reference template.

initTemplateRefBspline(self, ptIndex: int, et: float) → float¶

Returns:

A tuple containing:

• et

resetTracker(self) → None¶

Reset the tracker by freeing the memory allocated by the template tracker during the initialization.

setApprocHessian(self, approx: visp._visp.tt_mi.TemplateTrackerMI.HessienApproximationType) → None¶

setBlur(self, b: bool) → None¶

setBspline(self, newbs: visp._visp.tt_mi.TemplateTrackerMI.BsplineType) → None¶

setCostFunctionVerification(self, b: bool) → None¶

setCovarianceComputation(self, flag: bool) → None¶

setGain(self, g: float) → None¶

setGaussianFilterSize(self, new_taill: int) → None¶

setHDes(self, tH: visp._visp.core.Matrix) → None¶

setHessianComputation(self, type: visp._visp.tt_mi.TemplateTrackerMI.HessienType) → None¶

setIterationMax(self, n: int) → None¶

Set the maximum number of iteration of the estimation scheme.

Parameters:

n: int¶

Maximum number of iterations to stop the estimation scheme. A typical value is around 100.

setLambda(self, _l: float) → None¶

setMinimizationMethod(self, method: visp._visp.tt_mi.TemplateTrackerMIInverseCompositional.MinimizationTypeMIInverseCompositional) → None¶

setNbIterBrent(self, b: int) → None¶

setNc(self, newNc: int) → None¶

setPyramidal(self, nlevels: int = 2, level_to_stop: int = 1) → None¶

Set the number of pyramid levels used in the multi-resolution scheme. If nlevels > 1, the tracker uses a pyramidal approach.

Parameters:

nlevels: int = 2¶

Number of pyramid levels. Algorithm starts at level nlevels-1.

level_to_stop: int = 1¶

Last level of the pyramid that will be considered. Lowest level is zero.

setSampling(self, sample_i: int, sample_j: int) → None¶

Set the pixel sampling parameters along the rows and the columns.

Parameters:

sample_i: int¶

Sampling factor along the rows. If 1 all the lines are considered. If 2, consider one line over two.

sample_j: int¶

Sampling factor along the columns. If 1 all the columns are considered. If 2, consider one column over two.

setThresholdGradient(self, threshold: float) → None¶

setThresholdResidualDifference(self, threshold: float) → None¶

Set the threshold used to stop optimization loop. When the residual difference between two successive iterations becomes lower than the threshold we stop optimization loop.

Note

Increasing the default value allows to speed up the tracking.

Parameters:

threshold: float¶

Threshold used to stop optimization. Default value is set to 1e-4.

setUseBrent(self, b: bool) → None¶

By default Brent usage is disabled.

setUseTemplateSelect(self, b: bool) → None¶

Use only the strong gradient pixels to compute the Jabobian. By default this feature is disabled.

setp(self, tp: visp._visp.core.ColVector) → None¶

track(self, I: visp._visp.core.ImageGray) → None¶

Track the template on image I .

Parameters:

I: visp._visp.core.ImageGray¶

Image to process.

trackRobust(self, I: visp._visp.core.ImageGray) → None¶