An efficient and effective automatic image annotation using positive and negative example images

作者：Song Haiyu^*; Li Xiongfei; Wang Pengjie; Bao Cuizhu; Dong Yuying

来源：ICIC Express Letters, 2011, 5(8 B): 2927-2932.

• Annotation performance
• Automatic image annotation
• Classification approach
• Classification model
• Data sets
• Keyword vector classification
• Negative examples
• Probability model
• Time complexity
• Time cost
• Translation models

Automatic image annotation is an active topic and difficult task in computer vision domain, which has attracted more and more researchers' attention. Many approaches have been proposed to automatically annotate images. To improve the annotation quality, most of them have profound principlessophisticated modelsand high time complexities. How to efficiently annotate image is becoming one key issue in the field of computer vision. We propose a very efficient and effective approach called keyword vector classification approach, whose model is very simple. Experiments on Corel5K dataset show that the annotation performance of the our proposed approach is almost four times better than translation model and twice better than CMRM (in terms of precision, recall and F1) while the time cost of ours surprisingly decreases. In addition, to our knowledge, our proposed approach outperforms most of existing ones in terms of N+.

• 出版日期2011