Video Quality Estimation for Mobile Streaming Applications with Neuronal Networks
Michal Ries, Jan Kubanek, Markus Rupp · 2006
The provision of mobile multimedia streaming applications becomes essential for emerging 3G networks. The crucial point of successful deployment of multimedia mobile services is the user satisfaction level, since the perceptual video quality for such low bit rates, frame rates and resolutions is limited. Depending on the content character of a video sequence, the compression and network settings, maximizing the subjective perceptual quality also differs. The complexity of quality estimation and maximizing perceptual quality for mobile streaming application is still high if only the most significant influence factors are taken into account. Aim of this work is to design an artificial neural network with low complexity for the estimation of visual perceptual quality, based on a combination of a possibly small set of the most important objective parameters (compression settings and content features). To achieve this, the neural network was trained with a set of objective and subjective parameters, obtained by an extensive survey. Moreover, estimations with neural networks do not require any knowledge about the original sequence. The achieved correlation with the data set is as good as if the more the complex human vision based estimation is applied.