An active contour model for texture image segmentation using Rényi divergence measure

Abstract

This paper proposes an efficient method for active unsupervised texture segmentation. A new descriptor for texture features extractions based on Gaussian and mean curvature is constructed. Then the optimization of a functional who uses the R´enyi divergence measure and our descriptor is proposed in order to design an active contour model for texture segmentation. To get a global solution and efficient, fast algorithm, the optimization problem is redefined. The algorithm associated with this last optimization problem avoids local minimums and the run-time consuming compared to the level-set representation of our active contour model. In order to illustrate the performance of the technique, some results are presented showing the effectiveness and robustness of our approach.

Keywords

• level set theory
• texture segmentation
• global minimization
• differential geometry
• Rényi divergence measure
• shape optimization