Transrating-Assisted Spatial Resolution Conversion Methods For Compressed Videos

In pervasive media environments, users may access and interact with multimedia content on different types of networks and terminals with different screen sizes. A video adaptation tool or system is needed to generate a presentation for the video stream with a new resolution to meet user needs in customized situations. Spatial resolution transcoding is an effective technique responding to this challenge. In order to reduce the computation complexity, it is imperative for the transcoder to avoids motion estimation and mode decision partially by reusing the decoded motion vectors and macroblock modes. However, it is difficult to efficiently reuse the motion vectors and macroblock modes for both the pixel-domain scheme and the DCT-domain scheme because of the inconsistencies of the spatial resolution for video streams before and after transcoding, which leads to a sharp degradation of quality of the transcoded videos. Therefore, it is urgent to develop a new video spatial resolution conversion method, which can improve the video quality, as well as maintain a relatively lower computational complexity.A transrating-assisted spatial resolution conversion method (TARC) for compressed videos is proposed by utilizing the property of equivalence between downsampling of the video spatial resolution and bitrate reducing of the video streams in frequency domain. TARC carries out the conversion in a different way from the traditional spatial transcoding, which does not change the video spatial resolution before the conversion. Instead, it directly decodes the compressed video stream and then reencodes it to a new bitrate corresponding to the target video spatial resolution by reusing the motion vectors and the macroblock modes. That is, it firstly transcodes the original video stream to the version with a proper bitrate through a transrating process, while transferring the spatial resolution resampling to the receiving terminal to proceed before playback. A tactic which proactively sets the high frequency discrete cosine transform (DCT) coefficients to zeros for P frames is used in the transrating process to implicitly implement subsampling of the video while keeping its spatial resolution invariant, which facilitates efficient using of motion vectors and macroblock modes while reducing the computational load. Meanwhile, the resampled spatial resolution at the receiver matches the subsampled resolution of the transrating process in frequency domain, hence it also restrains the effect of aliasing perfectly caused by the resampling and keeps a good video quality.Experimental results show that it can significantly improve the rate distortion performance of the obtained video streams,providing a better video quality, about 2dB improvement in term of PSNR, compared to that obtained by the traditional spatial resolution transcoding methods, which shows widely promising applications for wireless video communications. It is also easy to implement spatial resolution conversion in arbitrary ratios, which is useful to support multimedia streaming adaption better. This thesis is supported by the project of National Natural Science Foundation of China (No.60873029), the project of Natural Science Foundation of Hubei province(No. 2010CDB01604) and the project of Innovation Research Foundation of Huazhong University of Science and Technology (No.2010MS014).