Reduced Fractional Modeling of 3D Video Streams: the FERMA Approach

Multimedia streaming of three-dimensional (3D) stereoscopic videos over last-generation networks subject to bandwidth limitations is an open problem. The development and spread of communication networks and devices that accept 3D videos is not supported by proper scheduling strategies. Namely the high variability of streams should be considered to reduce effects of network delays, packet losses, shortage of bandwidth resources, and shared use by multiple clients. Then, it is important to improve the characterization of 3D videos for more effective streaming. To this aim, this paper proposes a fractional exponential reduction moments approach based on the statistics of the so-called fractional moments. Each random sequence of frames in 3D videos can be analyzed and reduced to a finite set of parameters, that allow fitting to the sequence by exponential functions and then a characterization and classification of the video by a sort of fingerprint. The method does not depend on the format and the encoding technique of the video. Finally, the approach will allow comparing real streams and numerical data output from fractional dynamical models by means of the reduced parameters. Statistical proximity between time series and a fractional model or between different models simplifies formalization and classification of fractional models.