摘要 |
PROBLEM TO BE SOLVED: To provide a building image retrieval system which evaluates similarity of images belonging to a category of buildings with sufficient precision. SOLUTION: In the building image retrieval system for extracting feature values of the images of the buildings and calculating the similarity of the images to be compared based on the feature values, it is provided with a local area extraction means 36 for extracting similar areas based at least on colors, textures and shapes of the buildings and storing them as feature values of local areas, a global area extraction means 37 for specifying global areas by clustering the local areas based on area and the shape and storing the colors, the textures, the shapes and spatial relation between the global areas and the local areas as feature values of the global areas, an image data base 61 which stores each feature value of a plurality of original images of the buildings and an image retrieval means 39 for comparing each feature value of query images (exemplified images) with each feature value of the original images stored in the image data base and extracting the original images with high similarity. COPYRIGHT: (C)2006,JPO&NCIPI
|