Content-Based Adaptive Transmission for Soccer Video

摘要

To address the problem of how to achieve high quality on adapting user-interested content to the constraints of instantaneous network bandwidth, a content-based adaptive transmission system is proposed. Our main contributions include the following: User's preference on soccer events and exciting intensity implicated in affective content are combined to rank the importance of user-interested content in soccer video. The CBAT is formulated as 0/1 knapsack problem to employ a content-based frame dropping algorithm, which performs the real-time adaptation on multiple granularity of temporal content for optimal utility of perceptual quality and semantic quality. Experimental studies are conducted to compare the performance on network throughput, perceptual quality and semantic quality with the proposed solution, highlight-based adaptation and motion activity based adaptation. At last, the subjective tests demonstrate that our CBAT system achieves higher voting on three aspects of crispness, motion-smoothness and content visibility than the other methods.