Retrieval of images by spatial and object similarities |
| |
| Authors: | Shu-Ming Hsieh Chiun-Chieh Hsu |
| |
| Institution: | Department of Information Management, National Taiwan University of Science and Technology, Taipei 10607, Taiwan |
| |
| Abstract: | In the last several decades it has become an important basis to retrieve images from image databases (IDBs) by the semantic information held in the image objects and the spatial patterns formed by these objects. In this paper, we propose a new method for similarity retrieval of symbolic images by both the attributes and the spatial relationships of the contained objects. The proposed method CPM (common pattern method) retains the common spatial patterns of two images in new data structures CP_DAG (common pattern directed acyclic graph) and performs the similarity calculation efficiently in practice. The conducted experiments use both a synthetic dataset and an existing image database. The experimental results show that CPM outperforms LCS_Clique, SIMR, SIMDTC, and 2D Be-string for average efficiency and effectiveness. CPM also has steady efficiency while the number of image objects and the object symbol duplication rates increase. |
| |
| Keywords: | Image retrieval Similarity function Spatial relationship Common spatial pattern Directed acyclic graph |
| 本文献已被 ScienceDirect 等数据库收录！ |
