Akademska digitalna zbirka SLovenije - logo
ALL libraries (COBIB.SI union bibliographic/catalogue database)
  • Perception-oriented prominent region detection in video sequences
    Congyan, Lang ; De, Xu ; Xu, Yang
    Effective and efficient reprentation of video sequences is an important yet challenging task for video retrieval and browsing. In this paper, we propose a new approach for the prominent region ... detection from the viewpoint of the human perception intending to construct a good pattern for content representation of the video sequences. Firstly, we partition each frame into homogeneous regions using a technique based on a non-paramete clustering algorithm. Then, in order to automatically determine the prominent importance of the different homogenous regions in a fame, we extract a number of different mise-en-scene-based perceptual features which influence human visual attention. Finally, a modified Fuzy Inference Neural Networks is used to detect prominent regions in video sequence due to its simple structure and superior performance for automatic fuzzy rules extraction. The extracted prominent regions could be used as a good pattern to bridge semantic gap between low-level features and semantic understanding. Experimental results show the excellent performance of the approach.
    Source: Informatica : an international journal of computing and informatics. - ISSN 0350-5596 (Vol. 29, no. 3, 2005, str. 253-260)
    Type of material - article, component part
    Publish date - 2005
    Language - english
    COBISS.SI-ID - 19428391

Shelf entry