| This dissertation deals with research topics in the area of content-based access of image and video data. The main objective is to bridge the semantic gap between high-level concepts in the human mind and low-level features extractable by the machines. An emphasis is put on the learning and classification aspect during the interactive retrieval process, namely, relevance feedback algorithms. A novel algorithm, BiasMap, is proposed to take into account specifically the small sample asymmetric nature of the problem. A kernel and a boosting approach have been applied for achieving nonlinear capability. Other research efforts include local and global structural representations for images to capture more semantic information, the mixed use of textual and low-level features to facilitate intelligent access and user interaction, and content-based, nonlinearly sampled video delivery over low-bit-rate channels. |
