A feedback-based method for adaptive ROI protection in H.264/AVC

The bit error and packet loss in unreliable channels such as Internet and wireless networks lead the quality of video data unacceptable, which force us to take the network adaptive transmission into account. In this paper, we firstly propose a novel Greedy Spread (GS) ROI extraction method to extract ROI effectively, and then the network state changing is quickly perceived without occupying additional network overhead. According to the feedback about current network state from the receiver, Gilbert model is used to predict network packet loss, which enables the source encoder to dynamically adjust the ROI protecting schemes. Experimental results shows that the GS method extracts the ROI at high accuracy, also the adaptive method is of low computational complexity and high protection capability, suitable for real-time applications, so make a notable improvement in visual quality.