2D Partial Similarity Estimation Based on Energy Segmentation - I: Segmentation

How to perform partial similarity between models is still an open issue and attracts many attentions. Segmentation is regarded as one of the most critical and prospective steps for overcoming the task. An energy based segmentation approach for 2D curve is proposed in this paper. A curve is regarded as an elastic stripline(like spring) which has the balanced state and unbalanced state. Once it lies on the unbalanced state such as a spring elongated or compressed, it will possess the potential energy. Then the segmentation can be performed smoothly based on the energy and its attribute. Important is that no thresholds are required to initially determine the segmentation compared with the other methods. Also the proposed algorithm has several important and interesting properties such as robustness, independence,invariance and efficiency.