Using Hair-filament and Hair-strand Features for Hairstyle Classification

• Main Authors: Fei-hsiang Huang, 黃飛翔
• Other Authors: Jiann-shu Lee
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/62904021015980219600

碩士 === 國立臺南大學 === 資訊工程學系碩士班 === 99 === As the time changes, variety of hair styles expand, and it''s easy to obtain various kinds of information of hair styles; nevertheless, the larger the database is, the more important the hair classification become. Unfortunately, the structure of hair is very complicated, and the texture of hair is extremely unbalanced. The major problem is that there doesn’t exist an efficient way to classify those images. In this thesis we propose a method to manage the hair classifications and styles. We take the hair-filament orientation transition vector (HFOTV) and hair-stand orientation fourier descriptor (HSOFD) as hair features, then classify hair images by support vector machine (SVM). The experimental results show that the proposed method has extremely high classification accuracy.