Composition based semantic scene retrieval for ancient murals
PCM'10 Proceedings of the 11th Pacific Rim conference on Advances in multimedia information processing: Part I
| Hi-index | 0.00 |
This paper mainly deals with efficient retrieval of Dunhuang Frescoes from large image databases based on similarity calculation of color and texture features. In this paper, we firstly give out a survey of research on Dunhuang Frescoes by artists from all over the world. Therefore, these research achievements can provide us a new clue to organize and manage Dunhuang Frescoes databases, and guide us to find an efficient way to retrieve Dunhuang Frescoes from large-volume image databases. We present a similarity calculation method from multi-source analogy. Then we describe several techniques for extracting the color and texture features of Dunhuang Frescoes. Finally, we calculate the comprehensive similarity between Frescoes using color and texture features. By means of the similarity computation, we can carry on an efficient and fast image database retrieval. A Dunhuang Frescoes retrieval and analysis system demonstrates that retrieval using the combination of color and texture features can assure high precision and robustion.