With the technology advances in multimedia compression and Internet, video streaming of multimedia application is full of potentials. However, the delivery mechanism of the traditional Internet can not satisfied with the demand of multimedia application. Additionally, considering the video encoding process, the property of video packet with different types has different significance. An equal error protection to all video packets in the Internet will degrade the video quality significantly.

    Therefore, this thesis proposes a cross layer QoS mapping framework for H.264 video transmission over Diffserv network. Then there are three main parts in this framework; First, application layer proposes an adaptive priority of video packet mechanism. Furthermore, the mechanism can estimate each packet location and content information to adjust different video with adaptive priority. Second, network layer recommends a dynamic cross layer QoS mapping (D-CQM) mechanism. According to the current network situation, D-CQM can use the priority of video packet in application layer and it corresponds to the limited Diffserv level in network layer to regulate the dynamic QoS mapping. Third, for the extra packet header, it brings a packetization strategy that considers the priority of video packet and packet header overhead simultaneously. After evaluating the whole transmission efficiency, the cross layer QoS mapping framework over Diffserv network can improve the quality of video receiving and also better than the traditional Internet up to 4dB. Moreover, the adaptive priority of video packet is superior to the frame/packet based priority up to 3dB/1.5dB. For the strategy of priority packetization, it also increase PSNR up to 1.6dB than fixed packetization