2dLocalEstimators

Compares local estimators on implicit shapes using DGtal library.

Usage: 2dlocalEstimators –output <output> –shape <shapeName> [required parameters] –estimators <binaryWord> –properties <binaryWord>

Below are the different available families of estimators:

• True estimators
• Maximal DSS based estimators
• Maximal DCA based estimators
• Binomial convolver based estimators
• Integral Invariants based estimators

The i-th family of estimators is enabled if the i-th character of the binary word is not 0. The default binary word is '10000'. This means that the first family of estimators, ie. true estimators, is enabled, whereas the next ones are disabled. Below are the different available properties:

• Tangent
• Curvature

Allowed options are :

-h [ --help ]                      display this message
-l [ --list ]                      List all available shapes
-o [ --output ] arg                Output
-s [ --shape ] arg                 Shape name
-R [ --radius ] arg                Radius of the shape
-K [ --kernelradius ] arg (=0)     Radius of the convolution kernel (Integral
                                   invariants estimators)
--alpha arg (=0.33333333333333331) Alpha parameter for Integral Invariant 
                                   computation
-A [ --axis1 ] arg                 Half big axis of the shape (ellipse)
-a [ --axis2 ] arg                 Half small axis of the shape (ellipse)
-r [ --smallradius ] arg (=5)      Small radius of the shape
-v [ --varsmallradius ] arg (=5)   Variable small radius of the shape
-k [ --k ] arg (=3)                Number of branches or corners the shape
--phi arg (=0)                     Phase of the shape (in radian)
-w [ --width ] arg (=10)           Width of the shape
-p [ --power ] arg (=2)            Power of the metric (double)
-x [ --center_x ] arg (=0)         x-coordinate of the shape center (double)
-y [ --center_y ] arg (=0)         y-coordinate of the shape center (double)
-g [ --gridstep ] arg (=1)         Grid step for the digitization
-n [ --noise ] arg (=0)            Level of noise to perturb the shape
--properties arg (=11)             the i-th property is disabled iff there is
                                   a 0 at position i
-e [ --estimators ] arg (=10000)   the i-th estimator is disabled iff there 
                                   is a 0 at position i
-l [ --lambda ] arg (=0)           Use the shape to get a better 
                                   approximation of the surface (optional)
 -E [exportShape ] arf             Exports the contour of the source shape as a sequence of discrete points (.sdp)." )

Example: With this tool you can easely compare several estimator with the real value:

$  2dlocalEstimators --output curvature --shape flower --radius 15 -v 5  --gridstep 1  --estimators 11100 --properties 01

You can display the result by using gnuplot:

$ gnuplot
gnuplot> plot [][-0.4:0.35] 'curvature_True_curvature.dat' w lines title "true curvature" , 'curvature_MDCA_curvature.dat' w lines  title "Maximal DCA curvature estimator", 'curvature_MDSSl_curvature.dat' w lines title "Maximal DSS based estimators"

You should obtain such a graph:

Resulting visualization.

See also

2dLocalEstimators.cpp