| Description: |
Main mobile object localization is a task that emerges in research fields such us video understanding, object-based coding and various related applications, such as contentbased retrieval, remote surveillance and object recognition. The present work revisits mobile object localization in the context of content-based retrieval schemes and the related MPEG-7 framework, for natural and synthetic, indoor and outdoor sequences, when either a static or a mobile camera is utilized. The proposed multiresolution approach greatly improves the trade-off between accuracy and timeperformance leading to satisfactory results with a considerably low amount of computations. Moreover, based on the point gatherings extracted in [14], the bounding polygon and the direction of movement are estimated for each mobile object; thus yielding an adequate representation in the MPEG-7 sense. Finally, the resulting polygons can be used as appropriate initial estimates for methods that extract object contours, e.g. curve propagation approaches, such as [8,10] which utilize the level-set method. Experimental results over a number of distinct natural sequences have been included to illustrate the performance of the proposed approach. 1. |