透過您的圖書館登入
IP:18.191.234.62
  • 期刊

Enhancing Shape Retrieval Accuracy Using Possibilistic Fuzzy Alignment with Local Features

並列摘要


General approaches to object shape matching require a good correspondence between two sets of contour points of the object shapes. In this paper, we propose an effective and novel approach to help solving the problem with the concept of the possibilistic fuzzy alignment based on local features (PFALF). PFALF uses the positional information and the features generated by shape context (SC) or inner distance shape context (IDSC) for each pair of shape contour points. The score to match two shapes based on PFALF is designed and is further fused with that using dynamic programming (DC) to generate a new score. Finally we enhance the performance by an advanced affinity learning method. We prove that our framework is successful and competitive with known methods by experimenting on benchmark datasets: the Kimia-25, the MPEG-7 and Tari 1000.

延伸閱讀