An efficient motion vector composition approach is proposed to derive accurate initial motion vectors for the downsizing video during H.264/AVC downsizing transcoding. Our proposed algorithm uses the area-weighted non-zero AC coefficients to choose the motion vectors from the full-resolution videos as the initial motion vectors for the down-sized videos. Compared to other state-of-the-arts methods, our proposed achieves the best accuracy with the least computational complexity.
Multimedia applications such as video streaming and mobile TV are emerging as the most promising applications over wireless networks. The increased coding efficiency and network friendly architecture of the latest video coding standard H.264/AVC has facilitated the delivery of coded video content to wireless users. However, wireless networks allow lower transmission bit-rates than wired networks while the display resolution of mobile devices is generally smaller than that of standard definition (SD) TV. This calls for fast bit-rate reduction techniques through video transcoding that can deliver the best video quality to the mobile receiver while adhering to the bit-rate constraints of the wireless network. We present a bit-rate estimation model that speeds up the transcoding process by predicting the transcoded video bit-rate for different spatial resolution reduction ratios and quantization steps. We demonstrate that, on average, our proposed model can accurately estimate the bit-rate of the transcoded video to within 5% of the actual bit-rate of the transcoded video.
One objective in MPEG-2 to H.264/AVC transcoding is to improve the H.264/AVC compression ratio by using more advanced macroblock (MB) encoding modes. The motion re-estimation process is by far the most time consuming process in this type of video transcoding. In this paper, we present an efficient H.264/AVC block size partitioning prediction algorithm for MPEG-2 to H.264/AVC transcoding applications. Our algorithm uses rate-distortion optimization techniques and predicted initial motion vectors to estimate block size partitioning. It is also shown that using block size partitioning smaller than 8x8 (i.e., 8x4, 4x8 and 4x4) results in negligible compression improvements, and thus these sizes should be avoided in transcoding. .
Implementing MPEG-2 to H.264/AVC transcoding schemes in the pixel domain introduces a high degree of computational complexity. In the transform domain, this transcoding is more computationally efficient and several methods have been developed to address that approach. However, incompatibilities between the two standards such as the mismatches between the MPEG-2 and H.264/AVC motion compensation processes cause several distortions that may affect the overall picture quality. We address the main distortions that result from re-quantization errors, luminance half-pixel and chrominance quarter/three-quarter interpolation errors. New compensating algorithm that addresses the re-quantization errors and the interpolation errors (arising from the luminance as well as the chrominance components) is studied.
Since MPEG-2 and MPEG-4 both use 8x8 DCT in video coding, re-using the MPEG-2 DCT coefficients in MPEG-2 to MPEG-4 transcoding can significantly accelerate the overall transcoding speed. However, re-using MPEG-2 DCT coefficients brings quality distortions on the transcoded MPEG-4 videos. Through theoretical analysis, we conclude that the quality distortions are related to the following two errors: the re-quantization errors and the half-pixel chrominance motion compensation. Finally, algorithms are proposed to compensate for these quality distortions. As a result, the picture quality of the transcoded video is significantly improved: almost same as the picture quality resulting from the pixel-domain cascaded decoder-encoder transcoding structure. However, our algorithms requires much less computations.
In order to offer great flexibility and stability, the traditional video codec should be encapsulated as either static library or dynamic library to offer enough functionality for upper-lever applications. This project is to offer a baseline H.263 decoder that is encapsulated as dynamic libarary (Dll) for Windows applications. The encapsulation of the decoder uses the "Struct" concept in C programming language. At the same time, an enhanced error concealment algorithm is applied on the H.263 decoder. Our proposed H.263 decoder library was bought by a company in Tianjin, China.
Powered by Qiang Tang
© Copyright The University of
British Columbia, all rights reserved.