Presentation 2010-03-15
Performance Improvement of Bag-of-Keypoints by using Graph Cuts
Tetsuya FUJIKAWA, Motoaki SATO, Jiro KATTO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper, our goal is automatic segmentation of object regions which is processed before recognition in Bag-of-Keypoints. Region segmentation, such as Graph Cuts, is useful for object recognition. Objects tend to exist at the center of image, so we give seeds for a center and four corners of an image and apply the first-stage graph cuts to segment along edges. We get a color histogram from the first segmentation results and assign probabilities as the t-link of the graph for the second-stage graph cuts. Bag-of-Keypoints is then applied to the second segmentation results. We demonstrate that we can obtain better results than the case of using only Bag-of-Keypoints method.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Graph Cuts / Bag-of-Keypoints / Region Segmentation / Generic Object Recognition
Paper # IE2009-185
Date of Issue

Conference Information
Committee IE
Conference Date 2010/3/8(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Image Engineering (IE)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Performance Improvement of Bag-of-Keypoints by using Graph Cuts
Sub Title (in English)
Keyword(1) Graph Cuts
Keyword(2) Bag-of-Keypoints
Keyword(3) Region Segmentation
Keyword(4) Generic Object Recognition
1st Author's Name Tetsuya FUJIKAWA
1st Author's Affiliation School of Science and Engineering, Waseda University()
2nd Author's Name Motoaki SATO
2nd Author's Affiliation Graduate School of Fundamental Science and Engineering, Waseda University
3rd Author's Name Jiro KATTO
3rd Author's Affiliation School of Science and Engineering, Waseda University
Date 2010-03-15
Paper # IE2009-185
Volume (vol) vol.109
Number (no) 469
Page pp.pp.-
#Pages 6
Date of Issue