| Title |
Spin Image-based Relief Extraction from 3D Mesh of Stone Monuments |
| Authors |
박동렬(Dong-Ryul Park) ; 최강선(Kang-Sun Choi) |
| DOI |
https://doi.org/10.5573/ieie.2019.56.5.52 |
| Keywords |
메쉬 처리 ; 새김 인식 ; 3D 스캔 ; 문화 유산 |
| Abstract |
In this paper, we propose a relief extraction method using spin images. Spin images are well-known 3-D local surface descriptors. In the proposed method, we produce template spin images effectively representing reliefs, relief boundaries, and backgrounds. Then, the spin image for each vertex is compared with the three templates by measuring similarity, and is determined whether the vertex belongs to relief class or background class. In order to increase discriminative power, we found optimal parameters requiring for spin image generation. In addition, two stages of classification are applied to effectively reduce false positives. The proposed method outperforms conventional relief extraction methods by 0.14 in terms of F1 score. It was confirmed in the experiments that the proposed method can distinguish pecked noisy surfaces from reliefs. |