基於卷積神經網路書法字美學評估系統開發

Development of Calligraphy Character Aesthetics Evaluation System Based on Convolutional Neural Networks

杜依容(Yi-Rong Du)

指導教授 ： 吳孟倫

淡江大學/工學院/資訊工程學系碩士班/碩士(2022年)

https://doi.org/10.6846/TKU.2022.00112

摘要

漢字書法在藝術創作中有著不可抹滅的影響力，於美學角度而言，書法技法與美學更有著深遠的意涵，其重要性不言而喻。本研究提出一項書法字美學評估系統，利用卷積神經網路、Canny演算法與Shi-Tomasi角點偵測針對漢字書法進行審美評估。其美學宗旨在於提取書法字著重的筆劃、骨架和飛白三項構成審美的基本特徵，依據生成的楷書與隸屬兩種字體，個別分類為專業字體與四位不同使用者所書寫的業餘字體，客觀的探討筆墨線條間的書法藝術性。審美特徵搭配字體進行探討，概括地說，比對符合專業字體即為具藝術性的楷書或隸書，實驗結果平均約為91.9%，由此可論證此系統的有效性，並為使用者提供書法字體判別與辨識率，通過這項系統可提供客觀的書法字評價標準。

關鍵字

卷積神經網路 ； 書法 ； 美學評估 ； 審美特徵

並列摘要

Chinese calligraphy has an historical influence in artistic creation. From an aesthetic
point,technique of calligraphy and aesthetics have far-reaching implications, and their
importance is self-evident. This paper proposes a calligraphic character aesthetic
evaluation system,which uses convolutional neural network, Canny algorithm and
Harris corner detection to evaluate Chinese calligraphy. Its aesthetic purpose is to extract
the three basic characteristics of aesthetics, which are the strokes, skeleton and fei bai style
that calligraphy specialist focuses on. According to the generated regular script and official
script, they are individually classified into professional fonts, and amateur fonts written by
four different users. The objective discussion of calligraphy artistry between brush and ink
lines. Aesthetic characteristics are discussed with fonts. In a nutshell, the comparison with
professional fonts is an artistic regular script or official script. The average experimental
results are about 91.9%, which can prove the effectivenessof this system, and the
identification and recognition rate of calligraphy are provided forusers. Through this
system, objective evaluation criteria for calligraphy characters can be provided.

並列關鍵字

Convolutional Neural Networks ； Calligraphy ； Aesthetic evaluation ； Aesthetic features

參考文獻

[1] J. Zhang, W. Yu, Z. Wang, J. Li, and Z. Pan, “Attention-enhanced CNN for Chinese calligraphy styles classification,” in Proceedings of the IEEE 7th International Conference on Virtual Reality, Foshan, China, May 2021, pp. 352-358.

[2] Z. Ma and J. Su, “Aesthetics evaluation for robotic Chinese calligraphy,” IEEE Transactions on Cognitive and Developmental Systems, vol. 9, no. 1, pp. 80-90, March 2017. doi: 10.1109/TCDS.2016.2645598.

[3] L. Xiang, Y. Zhao, G. Dai, R. Gou, H. Zhang, and J. Shi, “The study of Chinese calligraphy font style based on edge-guided filter and convolutional neural network,” in Proceedings of the IEEE 5th International Conference on Signal and Image Processing, Nanjing, China, Oct 2020, pp. 883-887.

[4] J. Mano, L. He, T. Nakamura, H. Enowaki, A. Mutoh, and H. Itoh, “A method to generate writing-brush-style Japanese Hiragana character calligraphy,” in Proceedings of the IEEE International Conference on Multimedia Computing and Systems, Florence, Italy, June 1999, pp. 787-791.

[5] C. H. Chou, C. S. Wu and C. C. Han, “An interactive grading and learning system for Chinese calligraphy,” in Proceedings of the IEEE International Conference on Electro Information Technology, Tsukuba, Japan, June 2005, pp. 6.