Autocorrelation analysis of spatial features for mobile video services

The growing deployment of mobile video applications is calling for new methods to measure quality as perceived by humans. The suitable quality metrics may explore spatial and temporal video features that quantify the presence of related artifacts in an examined mobile video service. Considering a video as a sequence of images, computation of spatial features for each involved image would pose significant strain on system resources such as battery power and potentially induce large computational load in the mobile terminal. In this paper, we therefore examine the progression of spatial features in mobile videos over time using an autocorrelation approach. This enables us to reveal the duration over which spatial feature values of a mobile video may be considered as constant. In analogy to the characterization of mobile radio channels, we refer to this duration as the coherence time of an examined spatial feature for a given video. The provided numerical results illustrate that large reductions in the frequency of computing spatial features in mobile videos may be obtained. This in turn reduces the consumption of resources in the mobile terminal.