Digital Video Downsizing Algorithm Research

With the development of multimedia communication technologies, there is a high degree of convenience and enjoy the rich content of multimedia. However, the diversity of communication devices brings a new difficulties and challenges for the distribution, transmission, and consumes of multimedia. The diversity of multimedia contents also requires the terminal equipment has the ability to handling a wide range of multimedia formats. With the challenges of today's multimedia communication technology, it is hoped that an efficient form of multimedia accessing can be achieved. So Universal Multimedia Access (UMA) emerged. Universal Multimedia Access provides the ability to access rich content of multimedia in a heterogeneous network and terminal environments, at any time anywhere, fully considering the potential users'demands. And the digital video transcoding technology is the most complex, most of a system resource consumption technology to realization.In this paper, the basic theory of digital video transcoding and related standardization work are involved, and focus on the analysis and discussion on the algorithm to the spatial resolution reduction technology. As the macroblock coding type confliction on the existing frequency domain spatial resolution downscale technology, and the reference region error caused by the motion vector estimation, the frequency domain macroblock coding transform algorithm and the residual enhancement algorithm for the motion vector estimation are proposed. The macroblock coding transform algorithm extracts the reference region in DCT domain by matrix computation, which is pointed by motion vector. With the extracted region in DCT domain, the INTER macroblock can exchanged with INTRA macroblock, without further complicated decoding and re-coding in pixel domain. Residual enhancement algorithm based on motion vector estimation calculates the difference between original reference region and the estimated motion vector reference region, and fix the differences in residual for amendment, which avoids introducing new error in downscaling process and provides an significant improvement in transcoding quality in spatial resolution downscaling. Both algorithms are complements to current existing digital video transcoding technology, which process in the frequency domain and have a low computational complexity.A lot of experimental results show that the proposed algorithm can effectively solve the problems in the spatial resolution downscale transcoding, and further enhance the performance. In this paper, the proposed algorithm can be applied to other DCT-based video transcoding technology.