Kulczynski similarity index for objective evaluation of mesh segmentation algorithms

Recently, the 3D mesh segmentation is considered as an important stage in many applications in 3D shape analysis. In fact, extensive research has been performed to offer multiple approaches and algorithms for 3D mesh segmentation. Nevertheless, it is relatively hard to assess which algorithm produces more accurate segmentation quality than the other. Various methods have been proposed to evaluate the quality of segmentation algorithms. However, the evaluation of these qualities is far from being a solved problem, since most of the existing metrics don't take into account the regularity of the meshes and base their comparison on only one ground truth segmentation which presents a serious problematic since almost all the 3D objects have in general multiple reference segmentations. Hence, in this paper, we propose a new objective evaluation measure, based on the Kulczynski similarity index. The proposed method support comparison with multiple ground truth segmentations and takes into account the irregularity of meshes. Several experiments have been produced to show the discriminative power of our proposed measure.