Indexing flowers by color names using domain knowledge-driven segmentation

Abstract
We describe a solution to the problem of indexing imagesof flowers for searching a flower patents database by color.We use a natural language color classification derived fromthe ISCC-NBS color system and the X Window color namesto effectively use the domain knowledge available, provideperceptually correct retrieval and allow natural languagequeries. We have developed an automatic iterative segmentationalgorithm with knowledge-driven feedback to isolatea flower region from the background. The color of the floweris defined by the color names present in the flower regionand their relative proportions. The database can be queriedby example and by color names. We demonstrate the effectivenessof the strategy on a test database.

This publication has 8 references indexed in Scilit: