No-reference Quality of Experience estimation of H264/SVC stream

Scalable video coding (SVC) is considered as a promising video format for multimedia applications. It provides scalability in three-dimensional space: spatial scalability (picture size), temporal scalability (frame rate), and SNR/Quality scalability. The scalability is achieved by introducing a layered encoding of the bit-stream. The aim of the scalability is to provide the best and suitable video quality to customers with heterogeneous network conditions and terminals´ capabilities. Towards this, there is a profound need for an efficient evaluation of end-users perceived quality, called Quality of Experience (QoE). In this paper, we identify parameters, such as the video characteristics (e.g. resolution, bitrate...) and the network conditions and impairments (e.g. packet loss), which have a significant impact on SVC video quality. The proposed method estimates the end-user´s perceived quality, using a trained Neural Network. Subjective and objective test results were used for the training of the neural Network.