Image retrieval and classification based on MPEG-7 descriptors.
Date of Issue2008
School of Computer Engineering
In order to make use of the vast amount of multimedia data, e?cient and e?ective techniques to analyze and retrieve multimedia information based on its content need to be developed. In this thesis, we investigate three issues related to image retrieval and classi?cation based on MPEG-7 descriptors. Firstly, we propose a new similarity measure for one MPEG-7 color descriptor based on Earth Mover’s Distance (EMD). To reduce the computation time, M-tree index and lower bound of EMD are discussed, which can prune the images far from the query image. To combine two or more di?erent MPEG-7 descriptors to improve the retrieval performance, adescriptor-weighting scheme for combining multiple MPEG-7 visual descriptors is discussed. An optimization model is built to ?nd a set of optimal weights for a set of corresponding descriptors. Explicit solutions can be derived by Lag range multipliers, which are optimal and easy to calculate.
DRNTU::Engineering::Computer science and engineering::Information systems::Information storage and retrieval
Nanyang Technological University