Knowledge-Assisted Video Analysis for Content-Adaptive Coding and Transmission

Efficient video content management and exploitation, including coding and transmission, can greatly benefit from taking into account the semantics of the video content. However, the extraction of the latter is a non-trivial task that involves associating low-level features of the image domain with high-level semantic descriptions. In this paper, a knowledge-assisted approach for extracting semantics of domain-specific video content is presented and is employed for effecting content-adaptive coding and transmission of the video. To this end, domain knowledge considers both low-level features (color, motion, shape) and spatial behavior (topological and directional information) of video content for the purpose of analysis, as well as domain-, application- and user-specific importance factors associated with each domain concept that guide coding and transmission on the basis of the output of the analysis. Examples of the application of the proposed approach to sports videos, specifically belonging to the Formula One and Tennis domains, are provided in the Results section.