FLOWER IMAGE DATABASE CONSTRUCTION AND ITS RETRIEVAL
Shingo Nakamura, Mie Sawada, Yoshimitsu Aoki, Pitoyo Hartono, Shuji Hashimoto · 2001
Automatic object image retrieval is challenging task due to background complexity, variation of size and location, lighting condition, and so on. In such cases, additional constraints are required for the realization of effective and reliable retrieval system. In this paper, we propose a flower database contains of flower images (image database) and text data (text database). We also introduce a content-based retrieval method for flower images using both of textual keywords and a query image. Each flower in the database is indexed by using spatial image features (color, shape), and textual keywords that explain its characteristics (season, place, etc). The flower segmentation from background and color indexing are extended to extract the image features. By this indexing, our system provides flexible retrieval based on key images and keywords.