A novel tensor-information bottleneck method for multi-input single-output applications

Abstract

Ensuring timeliness and mobility for multimedia computing is a crucial task for wireless communication. Previous algorithms that utilize information channels, such as the information bottleneck method, have shown great performance and efficiency, which guarantees timeliness. However, such methods suit only in handling single variable tasks such as image processing, but are inapplicable to multivariable applications such as video processing. To address this critical shortcoming, we propose a novel tensor information channel which extends the current single-input single-output matrix information channel to a more practical multi-input single-output tensor information channel. In comparison with the classic information channel, our tensor information channel not only performs better in experiments, but also allows for a wider range of practical applications. We further build an innovative tensor-information bottleneck method upon the state-of-the-art information bottleneck method. Experiments on video shot boundary detection are conducted using benchmark data sets to demonstrate the effectiveness of our proposed approach compared with state-of-the-art methods. In specific, our approach yields a 6.2% increase compared with the information channel-based method, and when compared to other state-of-the-art methods, we achieve 0.1%–17.7% performance gains under different experimental configurations.