com.github.celldynamics.quimp.geom.filters

Class OutlineProcessor<T extends Shape<?>>

java.lang.Object

com.github.celldynamics.quimp.geom.filters.OutlineProcessor<T>

Type Parameters:

T - Outline or Snake class

public class OutlineProcessor<T extends Shape<?>>
extends Object

Support algorithms for processing outlines using their specific properties.

Author:

p.baniukiewicz

See Also:

ANA_, PointListProcessor

Constructor Summary

Constructors

Constructor and Description
OutlineProcessor(T outline) Assign outline to be processed to object.

Method Summary

Modifier and Type	Method and Description
double	amplificationFactor(double curv, double sigma, double magn) Return amplification factor dependent on curvature.
OutlineProcessor<T>	averageCurvature(double averdistance) Compute average curvature of Outline.
T	getO() Outline getter.
OutlineProcessor<T>	runningMean(int window, int iters) Apply running mean filter to Shape.
OutlineProcessor<T>	runningMedian(int window, int iters) Apply running median filter to Shape.
void	shrinkLin(double steps, double stepRes, double angleTh, double freezeTh) Shrink outline linearly.
void	shrinknl(double steps, double stepRes, double angleTh, double freezeTh, double sigma, double magn, double averageCurvDist, double averageNormalsDist) Shrink the outline nonlinearly.
OutlineProcessor<T>	sumCurvature(double averdistance) Sum smoothed curvature over a region of the membrane.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

OutlineProcessor

public OutlineProcessor(T outline)

Assign outline to be processed to object.

Parameters:

outline - Reference to Outline to be processed

Method Detail

runningMean

public OutlineProcessor<T> runningMean(int window,
                                       int iters)

Apply running mean filter to Shape.

Do not create new Shape but modify nodes of existing one. Compute Shape.calcCentroid(), Shape.updateNormals(boolean) and Shape.setPositions(). Normals are updated inwards.

Parameters:

window - size of filter window, must be uneven

iters - number of smoothing iterations

Returns:

reference to this object. Allows chaining

runningMedian

public OutlineProcessor<T> runningMedian(int window,
                                         int iters)

Apply running median filter to Shape.

Do not create new Shape but modify nodes of existing one. Compute Shape.calcCentroid(), Shape.updateNormals(boolean) and Shape.setPositions(). Normals are updated inwards.

Parameters:

window - size of filter window, must be uneven

iters - number of smoothing iterations

Returns:

reference to this object. Allows chaining

averageCurvature

public OutlineProcessor<T> averageCurvature(double averdistance)

Compute average curvature of Outline.

Does not apply to Snake. Require Outline.updateCurvature() to be called first to fill Vert.curvatureLocal field.

Parameters:

averdistance - Average distance in pixels to compute curvature. Nodes in half of distance on left and right from current node will be included in average curvature for this node.

Returns:

instance of this object.

sumCurvature

public OutlineProcessor<T> sumCurvature(double averdistance)

Sum smoothed curvature over a region of the membrane. For Outline only.

Does not apply to Snake. Require averageCurvature(double) to be called first to fill Vert.curvatureSmoothed field.

Parameters:

averdistance - Average distance in pixels to compute curvature. Nodes in half of distance on left and right from current node will be included in average curvature for this node.

Returns:

instance of this object.

shrinkLin

public void shrinkLin(double steps,
                      double stepRes,
                      double angleTh,
                      double freezeTh)

Shrink outline linearly.

Parameters:

steps - steps

stepRes - stepRes

angleTh - angleTh

freezeTh - freezeTh

See Also:

shrinknl(double, double, double, double, double, double, double, double)

shrinknl

public void shrinknl(double steps,
                     double stepRes,
                     double angleTh,
                     double freezeTh,
                     double sigma,
                     double magn,
                     double averageCurvDist,
                     double averageNormalsDist)

Shrink the outline nonlinearly.

This method uses Vert.curvatureSum computed by sumCurvature(double) to produce nonlinear scaling. Scaling factor depends on Vert.curvatureSum and it is computed by amplificationFactor(double, double, double) method. Nodes are shifted inwards according to normals (PointsList.getNormal()). If parameter averageNormalsDist is greater than zero, nodes within this distance to current node will be analysed to find lowest (must be negative) Vert.curvatureSum and then normal vector of this node will be copied to all other nodes in current window.

amplificationFactor(double, double, double) increase step size used by shrinkLin(double, double, double, double) maximally by factor of magn (amplificationFactor(double, double, double) option).

For linear shrinking use magn=1 and averageNormalsDist=0.

Parameters:

steps - number of steps - integer

stepRes - length of the step

angleTh - angle threshold

freezeTh - freeze threshold

sigma - sigma of Gaussian

magn - maximum amplification (for curv<<0)

averageCurvDist - number of neighbouring nodes to average curvature over. This is multiplier of average node distance. Always at least three nodes are averaged (n-1,n,n+1).

averageNormalsDist - number of neighbouring nodes to set normals the same (set 0 to disable).

See Also:

Outline.scaleOutline(double, double, double, double), amplificationFactor(double, double, double)

amplificationFactor

public double amplificationFactor(double curv,
                                  double sigma,
                                  double magn)

Return amplification factor dependent on curvature.

For positive curvature it returns 1, for negative a reversed Gaussian curve scaled between 1.0 and specified magnitude is returned.

 
 x=linspace(-2,0,100);
 sig = 0.3;
 u = 0;
 a = 4
 y = (a*(1-exp(-0.5*((x-u)/sig).^2)))/(a/(a-1))+1;
 plot(x,y)
 
 

Parameters:

curv - curvature value to compute magnification coefficient for

sigma - sigma of Gaussian

magn - maximum amplification (for curv<<0). Set to 1.0 for constant output of 1.0

Returns:

amplification coefficient

getO

public T getO()

Outline getter.

Returns:

the outline