No abstract

No abstract

Multimedia quality assessment has become an important issue for both monitoring and control of multimedia networks and for the optimization of the new algorithms for audiovisual enhancement. In this paper we evaluate various objective video quality metrics in respect to their match to the subjective assessment represented as mean opinion score (MOS). The proposed framework is aimed for low bit-rate MPEG-2 sequences in SD and HD format. We consider 29 objective measures for no-reference video quality assessment and rank them according to their match to the MOS independently and based on the cross-correlation between the objective measures. The selected set of ranked measures are incorporated into the proposed neural network scheme for estimating the overall visual perception. The selection of measures is done using the proposed linear model which aims at maximizing measure correlation to MOS and at the same time minimizing mutual correlation between the selected set of measures. The proposed VQA model is intended for (i) estimation of the overall visual quality and (ii) the evaluation of the objective video quality metrics for particular applications defined by the test sequences and corresponding visual perception results. We present the results of the proposed method on SD and HD sequences for different number of selected measures and provide comparison results (to the state-of-the-art methods) of the overall estimated quality.Index Terms-video quality assessment, feature selection, neural networks.

No abstract

No abstract