| Title |
A Study on the Generalization Method of Segmented Inscription Recognition Index to Improve 3-D Inscription Recognition Quality |
| Authors |
정범채(Beom-Chae Jeong) ; 최예찬(Ye-Chan Choi) ; 셰리프 물탈라(Sheriff Murtala) ; 최강선(Kang-Sun Choi) |
| DOI |
https://doi.org/10.5573/ieie.2021.58.6.79 |
| Keywords |
Text Segmentation; Inscription Extraction; Evaluation Metrics |
| Abstract |
This paper proposes an enhanced measure indicating the subjective quality of how recognizable the inscriptions extracted from 3-D steles are by generalizing the segmented inscription recognition index (SIRI). In inscription extraction, recognition quality of the inscription is related to not only a quantitative accuracy but also detected parts of the inscription. The SIRI evaluates the quality of inscription segmentation by partitioning 3-D data into four regions and assigning different regional importance to them accordingly. In the proposed method, the 3-D data are subdivided further finely using the levelset function to assign regional importance elaborately. As a result, the proposed measure can more accurately represent the subjective assessment of the extracted inscription compared to the SIRI. The experimental results showed that the proposed measure was more similar to the subjective inscription recognition than the conventional measures. |