Frame distortion estimation for unequal error protection methods in scalable video coding (SVC)

•We propose an accurate low complexity utility estimation technique that can be used in different applications.•This technique estimates the utility of each network abstraction layer (NAL) by considering the error propagation to future frames.•We utilize this method in an UEP framework with the scalable extension of H.264/AVC codec.•It achieves almost the same performance as highly complex estimation techniques (an average loss of 0.05 dB).•We propose a low delay version of this technique that can be used in delay constraint application.

In many multimedia applications, coded video is transmitted over error prone heterogeneous networks. Because of the predictive mechanism used in video coding, transmission error would propagate temporally and spatially and would result in significant quality losses. In order to address this problem, different error resilience methods have been proposed. One of the techniques, which is commonly used in video streaming, is unequal error protection (UEP) of scalable video coding (SVC). In this technique, different independent layers of an SVC stream are protected differently and based on their importance by using forward error correction (FEC) codes. Accurately analyzing the importance or utility of each video part is a critical component and would lead to a better protection and higher quality of the received video. Calculation of the utility is usually based on multiple decoding of sub-bitstreams and is highly computationally complex. In this work, we propose an accurate low complexity utility estimation technique that can be used in different applications. This technique estimates the utility of each network abstraction layer (NAL) by considering the error propagation to future frames. We utilize this method in an UEP framework with the scalable extension of H.264/AVC codec and it achieves almost the same performance as highly complex estimation techniques (an average loss of 0.05 dB). Furthermore, we propose a low delay version of this technique that can be used in delay constrained application. The estimation accuracy and performance of our proposed technique are studied extensively.