Authors:
D. G. Fernández-Pacheco
1
;
F. Albert
2
;
N. Aleixos
2
;
J. Conesa
1
and
M. Contero
2
Affiliations:
1
Universidad Politécnica de Cartagena, Spain
;
2
Universidad Politécnica de Valencia, Spain
Keyword(s):
Sketching, Segmentation, Simulated annealing.
Related
Ontology
Subjects/Areas/Topics:
Computer Vision, Visualization and Computer Graphics
;
Geometry and Modeling
;
Interactive Environments
;
Sketch-Based Interfaces
;
Sketch-Based Modeling
Abstract:
One of the main problems in the segmentation of freehand sketches is the difficulty of tuning the parameters involved in the process. Commonly, these parameters are chosen empirically from the observation of segmentation results in training sets. However, this approach rarely gets the best set of parameters, especially when the parameters depend on each other. This work presents an optimization algorithm, based on the simulated annealing technique, which tunes the segmentation parameters to improve segmentation results. The tuning of parameters has been formulated as an optimization problem where the cost function is expressed as the number of errors in the segmentation of a training set. Errors are determined comparing the computer segmentation with the correct one defined during the design of the shapes of the training set. Experimental work used 177 samples of 20 different shapes, achieving a performance ratio of 97.0% for the correct segmentation after tuning of parameters.