AnnoSearch: Image Auto-Annotation by Search
CVPR '06 Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition - Volume 2
Correlated Label Propagation with Application to Multi-label Learning
CVPR '06 Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition - Volume 2
Annotating Images by Mining Image Search Results
IEEE Transactions on Pattern Analysis and Machine Intelligence
A New Baseline for Image Annotation
ECCV '08 Proceedings of the 10th European Conference on Computer Vision: Part III
Proceedings of the 18th international conference on World wide web
NUS-WIDE: a real-world web image database from National University of Singapore
Proceedings of the ACM International Conference on Image and Video Retrieval
Learning social tag relevance by neighbor voting
IEEE Transactions on Multimedia
Multi-label feature transform for image classifications
ECCV'10 Proceedings of the 11th European conference on Computer vision: Part IV
Hi-index | 0.00 |
Image annotation can significantly facilitate web image search and organization. Although it has been studied for years by the computer vision and machine learning communities, image annotation is still far from practical. Existing example-based methods are usually developed based on label co-occurrence information. However, due to the neglect of the associated label set's internal correlation and relevance to image, the annotation results of previous methods often suffer from the problem of label ambiguity and noise, which limits the effectiveness of these labels in search and other applications. To solve the above problems, a novel model-free web image annotation approach is proposed in this paper, which consider both the relevance and correlation of the assigned label set. First, measures that can estimate the label set relevance and internal correlation are designed. Then, according to the above calculations, both factors are formulated into an optimization framework, and a search algorithm is proposed to find a label set as the final result, which reaches a reasonable trade-off between the relevance and internal correlation. Experimental results on benchmark web image data set show the effectiveness and efficiency of the proposed algorithm.