Evaluation and fusion of image segmentation methods

In this paper, we propose an evaluation method for regions-segmentation algorithms. This method was then applied on seven well-selected existing techniques. This allows us to deduce the interest of merging two techniques. In fact, we combine an hierarchical approach of regions segmentation by adaptive thresholding (HSA) in order to extract the uniform area, with a classification approach according to fuzzy c-means algorithm (FCM) looking for the extraction of textured areas. The robustness of the introduced approach was illustrated while applying it on several real images belonging to five different fields. The experiments and the evaluations showed a good quality independently of the application domain (presence of the uniform zones and/or textured zones in the treated image)