Skip to Main Content
Video databases are very demanding systems as far as mass storage requirements and computational resources necessary to perform common database operations, such as browsing and retrieval, are required. These operations can be simplified both in terms of the computational complexity and of processing time by performing them on an ensemble of frames, called key frames, representative of the content units (shots) in which a video can be segmented. An adaptive key frames extraction method based on a wavelet based multiresolution analysis in a perceptually uniform color space is presented. Experimental results that show the effectiveness of the proposed technique to select key frames summarizing the video's content, are finally provided.