Texture Orientation Estimation Based on Phase Congruency and Support Vector Machines

摘要：

This paper presents a novel approach to estimate texture orientations. The proposed approach benefits from the fact that texture orientations can be analyzed by micro-edge information. First, sample images were divided into a number of blocks. Second, feature vectors were generated from slopes of edges which presented in every block. These edges were obtained from phase congruency estimation. Third, a SVM model was trained using these features comprising both synthetic texture generated with controlled orientation and real texture images rotated at angles with a 10 degree interval. Differing from existing methods, this approach provides an effective method to analysis real textures without requiring a large set of training data. Experimental results show the effectiveness and accuracy of our proposed approach.

关键词： component phase congruency orientation SVM

作者: Ping Ji Junyu Dong Mike Chantler Muwei Jian Peng Chen

作者单位: Ocean University of China Qingdao, China Heriot-Watt University Edinburgh, UK The Hong Kong Polytechnic University Hung Horn, Kowloon,Hong Kong Shandong Entry-Exit Inspection and Quarantine Bureau

会议类型: 国际会议

会议名称: The 2012 International Conference of Agricultural Engineering and Food Engineering(2012年国际农业工程与食品工程学术会议 ICAE 2012)

会议地点: 哈尔滨

会议语种:英文

页码: 553-556

在线出版日期: 2012-06-16（万方平台首次上网日期，不代表论文的发表时间）