The Design and Use of Steerable Filters
IEEE Transactions on Pattern Analysis and Machine Intelligence
Shape Matching and Object Recognition Using Shape Contexts
IEEE Transactions on Pattern Analysis and Machine Intelligence
Object Recognition from Local Scale-Invariant Features
ICCV '99 Proceedings of the International Conference on Computer Vision-Volume 2 - Volume 2
Scale & Affine Invariant Interest Point Detectors
International Journal of Computer Vision
A Sparse Texture Representation Using Local Affine Regions
IEEE Transactions on Pattern Analysis and Machine Intelligence
A Performance Evaluation of Local Descriptors
IEEE Transactions on Pattern Analysis and Machine Intelligence
A Maximum Entropy Framework for Part-Based Texture and Object Recognition
ICCV '05 Proceedings of the Tenth IEEE International Conference on Computer Vision (ICCV'05) Volume 1 - Volume 01
Deformation Invariant Image Matching
ICCV '05 Proceedings of the Tenth IEEE International Conference on Computer Vision - Volume 2
Object Categorization by Learned Universal Visual Dictionary
ICCV '05 Proceedings of the Tenth IEEE International Conference on Computer Vision - Volume 2
International Journal of Computer Vision
SURF: speeded up robust features
ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part I
Adapted vocabularies for generic visual categorization
ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part IV
Sampling strategies for bag-of-features image classification
ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part IV
| Hi-index | 0.00 |
Bag-of-words is the state-of-the-art method used in visual categorization. The performance of visual categorization depends on four main operations: the detection of interest point, the description of interest point, the design of classifier, and the construction of codebook. In this paper, we focus on the optimizations of the first three operations. Firstly, we compare several popular detectors of interest points and propose an optimal detector combined MSER detector with Hessian-Laplace detector to sample the key points. This detector well combines the interest region with the interest point such that the image can be represented in a hierarchical way. Secondly, we adopt SIFT to describe the sampling region because our experiment results demonstrate that SIFT is more robust than other popular descriptors. Thirdly, we use SVM with RBF kernel for object classification. The proposed classifier outperforms other classifier in terms of the classification accuracy. In order to verify three proposed optimal operations, we implement them in two image datasets: Caltech and KTH-TIPS. The experimental results show that our optimal operations can increase the accuracy of object categorization.