원문정보
초록
영어
Image retrieval methods have been significantly developed in the last decade. The BOW (Bag-of-words) model lacks spatial information. Some methods stem from BOW approach which is recently extended to a vector aggregation model. Most of them are either too strict or too loose so that they are only effective in limited cases. In this study, we present a novel feature extraction method for image retrieval. We acquire the gradients features from the p.d.f (Probability density function) because of essentially representing the image. We construct the features by the histogram of the oriented p.d.f gradients via aggregation of the orientation codes. Then, we adopt the PCA (Principal component analysis) method to reduce the dimensionality of BOW. Furthermore, we introduce a novel and robust re-ranking method with the k-nearest neighbors. We estimate our method using various datasets. In the experiments on scene retrieval, the proposed method is efficient, and exhibits superior performances compared to the other existing methods.
목차
1. Introduction
2. Summary of the Related Works
3. Proposed Methods
3.1 Oriented Probability Density Function Gradients
3.2 Principal Component Analysis
3.3 Aggregation of p.d.f Gradient Orientation Codes
3.4 k-NN Re-Ranking
4. Experiments and Analysis
4.1 Results of k-NN Re-Ranking
4.2 Comparisons to Other Methods
4.3 Scalability for Large Datasets
5. Conclusions
Acknowledgments
References