Part-Based Shape Retrieval with Relevance Feedback
2005 IEEE International Conference on Multimedia and Expo
We introduce a relevance feedback mechanism for part-based shape retrieval. The database polygons are decomposed but the query polygon is not. In an initial search in the database, the database polygon parts are matched against the query polygon. The best matches are shown to the user, who has to decide which are relevant to its query. In successive iterations, the system infers which parts of the query are of interest, and makes a search with those parts only. . INTRODUCTION The retrieval
... The retrieval problem we consider is the following: given a large collection of images of polygonal shapes and a query polygon, we want to retrieve those shapes in the collection that "share" some parts with the query polygon. A possible approach is to formulate a query by selecting a part, or a combination of parts, from a polygon. The query must then be tested for similarity against the polygons in the database, and the best results are presented to the user. The retrieval performance depends on the selection of parts comprising a query, and the query process itself is labour intensitive. A natural extension to this type of retrieval is to relieve the user from the identifying the parts with good discriminative power. One way to do this is to divide the query polygon into parts, and to match all these parts and the possible combinations against the database polygons. An alternative is to decompose the database polygons and to match their parts against the query polygon. The larger the number of parts in the polygon decomposition, the smaller the number of similarities that remain undetected (false negatives) in the retrieval process. But since the number of possible combinations of parts of a polygon is exponential in the number of parts, a larger number of parts in the query or database polygon increases the query response time. Moreover, not all parts have the same discriminative power, and some combinations of parts may be irrelevant, thus it would be undesirable to search the database with them. To avoid these problems, we propose a relevance feedback mechanism for part-based shape retrieval. In our approach, the database polygons are decomposed but the query This research was supported by FP6 IST Network of Excellence 506766 AIM@SHAPE.