L2SSP: Robust keypoint description using local second-order statistics with soft-pooling

详细信息    查看全文

• 作者：Tiecheng Song^a ; ^{tggwin@gmail.com} ; ^{stc1984@126.com}Author Vitae ; Fanman Meng^bAuthor Vitae ; Qingbo Wu^bAuthor Vitae ; Bing Luo^bAuthor Vitae ; Tianqi Zhang^aAuthor Vitae ; Yongjun Xu^aAuthor Vitae
• 关键词：Local descriptor ; Keypoint ; Image feature ; Image matching ; Feature description ; Histograms ; SIFT ; LBP ; Covariance
• 刊名：Neurocomputing
• 出版年：2017
• 出版时间：22 March 2017
• 年：2017
• 卷：230
• 期：Complete
• 页码：230-242
• 全文大小：2880 K
• 卷排序：230

文摘

We model each pixel as local spatial-frequency patterns and local extremum patterns. We propose a soft spatial binning to adaptively aggregate the generated feature set. We represent keypoints as the second-order statistics to capture feature correlations. Our descriptor is rotationally invariant, highly discriminative, and robust to noise.