New evaluation method using sampling theory to evaluate 3D segmentation algorithms

3D segmentation and its performance evaluation play a crucial role in computer vision. Due to its importance, much effort has been consecrated to the segmentation process in the last decades. Consequently, the development of reasonable criteria for evaluating and comparing the performance of segmentation algorithms has become a major challenge in the area. In this work, we propose a new method for evaluating 3D segmentation algorithms. Our proposed approach is based on sampling theory to calculate a score of dissimilarity between an automatic and ground truth segmentations. Real experiments demonstrate the usability and efficiency of the proposed approach.