Automatic Image Annotation Based on Region Feature - Multimedia and Signal Processing - page 144

precision recall F1

0.5

0.4

0.3

0.2

0.1

0

1 2 3 4 5

Fig. 2. Performance with different values of annotation m

5 Conclusions and Future Work

We propose an automatic image annotation method based on region feature related.
General processes are: segmenting the image, extracting visual features of each
region, building the connection between image region and semantic concept, then
figuring out the joint probability in between, make clear the semantic by using
keywords relating method and considering the similarity of region features, finally
fulfill the automatic image region semantic annotation. Experiments show that the
proposed algorithm will help to get much better recall and precision.

Acknowledgment. This work was supported by the Innovation Team Grant of
SuZhou Vocational University NO.3100124 and the Open Foundation of SuZhou
Vocational University No. 2011SZDYJ07.

References

1. Jeon, J., Lavrenko, V., Manmatha, R.: Automatic Image Annotation and Retrieval Using
Cross-media Relevance Models. In: Proc. ACM SIGIR, pp. 119-126 (2003)
2. Lavrenko, V., Manmatha, R., Jeon, J.: A Model for Learning the Semantics of Pictures. In:
Proc. NIPS (2003)
3. Feng, S.L., Manmatha, R., Lavrenko, V.: Multiple Bernoulli relevance models for image
and video annotation. Trans. Circuits and Systems for Video Technology 13(1), 26-38
(2003)
4. Cusano, C., Ciocca, G., Schettini, R.: Image annotation using SVM. In: Proc. Internet
Imaging, pp. 330-338 (2004)
5. Carneiro, G., Vasconcelos, N.: A database centric view of semantic image annotation and
retrieval. In: Proc. ACM SIGIR, pp. 559-566 (2005)
6. Yang, C.B., Dong, M., Hua, J.: Region-based Image Annotation using Asymmetrical
Support Vector Machine-based Multiple-Instance Learning. In: Proc. CVPR, pp. 2057-
2063 (2006)