No-reference Quality Assessment For The Networked Video

To provide satisfying video services for consumers over time-varying networks, it is necessary to monitor the quality of networked video in real-time at network nodes. Accordingly, no-reference video quality assessment based on the parametric packet-layer model is addressed in this paper. The parametric packet-layer model uses the packet-header information without media-related payload information, so the computational complexity of this method is very light. The proposed no-reference method is very suitable for quality assessment at network nodes for in-service non-intrusive monitoring.The key to this no-reference, parametric packet-layer model is how to obtain useful information using the packet-header information and how to assess the quality of video stream using the obtained information. It is demonstrated through analyzing the experimental results that characteristics of video content are crucial for video quality assessment using the bit-rate, where the bit-rate for coding frames with different types could represent the characteristic of video content.Through analyzing the distribution of the bit-rate for coding frames with different types, two methods are proposed in this paper to detect the I frames in the video stream. Considering advantages and disadvantages of each method, a more advanced detection method is proposed by integrating both methods. Experimental results show that the integrated method is more accurate in I frame detection.Based on the proposed integrated method for I frame detection, a no-reference quality assessment measure for networked video without considering packet loss is introduced in the paper through studying the relationship of characteristics of video content and the ratio of coding bit-rate to the bit-rate for coding I frames. The experimental results have demonstrated the effectiveness of the proposed no-reference video quality assessment measure.