Title :
Object retrieval in image databases using image composition
Author :
Philipp-Foliguet, Sylvie ; Lekkat, Mohamed
Author_Institution :
ENSEA, UCP, Cergy, France
Abstract :
We propose an algorithm of image retrieval from a partial request. In order to retrieve images containing a specific object or group of objects, we use an algorithm of inexact graph matching, adapted to adjacency graphs of regions. The number of regions representing the object can be different from an image to the other. The algorithm uses a search tree in which only similar nodes are developed. The adjacency and spatial relationship of regions are used to prune the tree and accelerate the retrieval. The algorithm computes a distance between graphs, allowing to rank images. This distance takes into account both the distance between region features and the relative position of regions.
Keywords :
graph theory; image matching; image retrieval; image segmentation; tree searching; visual databases; fuzzy regions; graph matching algorithm; image composition; image databases; image retrieval; object retrieval; region-based retrieval; search tree pruning; spatial relationship; Acceleration; Image databases; Image retrieval; Image segmentation; Information retrieval; Spatial databases; Spatial resolution; Topology; Tree graphs; Visual system;
Conference_Titel :
Computer Graphics and Image Processing, 2003. SIBGRAPI 2003. XVI Brazilian Symposium on
Print_ISBN :
0-7695-2032-4
DOI :
10.1109/SIBGRA.2003.1241004