The Research On Stereo Video Codec For Mobile Terminals

Nowdays, the mobile phone has become one of the most important carriers for multimedia spread. With the development of the display devices and the wireless communication technology, mobile TV is believed to be an attractive market for the introduction of stereoscopic 3D. In order to remove redundant information, stereo videos based on the mobile terminals need more efficient video compression technology. As a new generation of international coding standard, H.264 has higher compression efficiency, a better adaptability to channel, as well as better error concealment capability, so H.264 can be used for video compression of mobile terminal. Considering the characteristics of mobile terminals, which includes low computational complexity, only two-channels for display and the limited bandwidth of wireless networks, the stereo video codec based on H.264 has been studied.Firstly, the paper compares the current stereo video coding methods, which include dual-channel simulcast coding, the joint video coding, the asymmetric resolution coding, as well as the left point view plus disparity coding, and the rate-distortion performances are analysed to appraise which is the best choice to be used for mobile terminals. The experimental results show that the joint video coding works well at high bit-rate; while in the low bit-rate, the asymmetric resolution coding not only maintains good subjective quality, but also saves bitrate of network bandwidth, so it is a better choice for stereo coding.Then, we find that sampling direction has close relation with rate-distortion performance in the asymmetric resolution coding scheme through experiments. The reason of this phenomenon is disscussed and analysed. At the same time, a new scheme is proposed to prejudge the down-sampling direction, which takes advantage of the horizontal prediction and vertical prediction information with 16×16 macroblock of left channel in intra code. The experimental results show that the proposed scheme can obtain a better performance without increase of complexity. And another work in the paper is on the decoder of the asymmetric resolution coding with disparity compensation prediction which is raised by Fehn. In the original structure, the decoder not only need to decode two channels, but also reconstruct the images of right channel. In this paper, a new scheme is proposed, which takes advantage of SKIP mode of macroblock to resize the resolution. It maintains the reconstruction PSNR value, and reduces the computational complexity of the up-sampling. Finally, considering the requirements of stereo video coding for real-time performance, a video coding framework based on mobile terminal is proposed. And the corresponding fast algorithm is aslo proposed on the framework. The algorithm takes that the time has stronger correlation than view into account in the stereo video coding, and quickly selects the proper reference frame and macroblock mode to achieve lower computational complexity, which reduces the complexity of stereo coding obviously.