Complex Printed Uyghur Document Image Retrieval Based on Modified SURF Features
As an important part of information retrieval,it is important to improve the accuracy of document image retrieval system.This paper proposes a document image retrieval method based on modified SURF features.Firstly,FAST+SURF features are extracted from the image,and then the similarity degree is retrieved by using different kinds of distances and matching points respectively.With the change of size,angle and illumination,the FLANN bidirectional matching and KD-Tree+BBF matching are implemented for its feature points; finally,based on these two kinds of retrieval methods,various Uyghur document image databases that have been collected and retrieved are searched.The experimental results indicated that both search methods can achieve accurate search requirements,but in computational complexity based on the matching number of retrieval is more convenient.At the same time,the comparison experiment proves that the proposed method is superior to the original feature in the retrieval time.
SURF feature FALNN bidirectional match KD-Tree and BBF match Complex document image retrieval
Aliya Batur Patigul Mamat Wenjie Zhou Yali Zhu Kurban Ubul
School of Information Science and Engineering,Xinjiang University,Urumqi 830046,China School of Mathematics and Information,Hotan Normal College,Hotan 848000,China
国际会议
广州
英文
99-111
2018-11-23（万方平台首次上网日期，不代表论文的发表时间）