MULTIINDEX QUERY ON IMAGE DATABASE

• Main Author: Saladin, Ahmad
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/78293
• Institution: Institut Teknologi Bandung

When searching for a specific image in a database, user may need to use various features. Various features are obtained by various feature extraction methods and need to be stored in multiple indexes. Currently, multimedia databases such as Cottontail DB support multiple indexes obtained from various feature extraction methods but image retrieval can only be done on one index. In this final project, an image retrieval system with query that facilitates searching on multiple indexes in the database is built so that users can convey their needs regarding an image they are looking for. In this system, to build indexes, the system provides several feature extraction methods. More feature extraction methods can be added by futere developer. To facilitate the user's image requirements, the system accepts a query representation consisting of predicates that describe the user's needs. The predicates are separated by the and, or, or, not operator. In the query representation, retrieval will be carried out for each predicate by utilizing an index that corresponds to the features of the predicate. Results on different indices are combined by calculating a final similarity value using fuzzy logic according to the operators used between predicates. Testing on the system shows that the system can facilitate retrieval on different indexes and the query representation received by the system can facilitate the needs of users.