Video Google: A Text Retrieval Approach to Object Matching in Videos
ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
Distinctive Image Features from Scale-Invariant Keypoints
International Journal of Computer Vision
Scalable Recognition with a Vocabulary Tree
CVPR '06 Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition - Volume 2
Kernel Codebooks for Scene Categorization
ECCV '08 Proceedings of the 10th European Conference on Computer Vision: Part III
Hamming Embedding and Weak Geometric Consistency for Large Scale Image Search
ECCV '08 Proceedings of the 10th European Conference on Computer Vision: Part I
Improving Bag-of-Features for Large Scale Image Search
International Journal of Computer Vision
IEEE Transactions on Pattern Analysis and Machine Intelligence
Product Quantization for Nearest Neighbor Search
IEEE Transactions on Pattern Analysis and Machine Intelligence
In defense of soft-assignment coding
ICCV '11 Proceedings of the 2011 International Conference on Computer Vision
| Hi-index | 0.00 |
Image representation is essential to performance of content-based image retrieval. VLAD has been proved to be superior to BOF. However, hard assignment is utilized in VLAD, which does not consider codeword uncertainty and codeword plausibility. In this paper, each cluster associated to visual word is defined as a hyper-sphere. The radius is denoted as the distance from visual word to the farthest feature point. Spherical soft assignment is proposed to adaptively assign a local feature to close visual words according to corresponding radius. Spherical soft assignment and a descriptor-space soft assignment of state of the art are applied to VLAD. Experiments on multiple datasets demonstrate that the proposed spherical soft assignment can noticeably improve VLAD image representation in image retrieval and be superior to the descriptor-space soft assignment.