视频会议中快速视频编码器的研究与实现

RESEARCH AND IMPLEMENTATION OF FAST VIDEO CODER FOR VIDEOCONFERENCE

动态图象的快速编码是视频会议系统实现的关键问题之一 .本文在快速 DCT和快速运动估计算法的基础上 ,融合两种工程性的解决方法 :提前滤除全 0 DCT系数块和基于 MMX的并行计算 .在不增加输出码率和不影响图象质量的情况下大幅度地提高了视频编码器的编码速度 .实验结果表明 ,所实现的视频编码器可充分满足 IP视频会议。

Fast encoding motive image is one of the key techniques for implementing videoconference system. Based on a fast DCT algorithm and a fast motion estimation algorithm used in video coding, two engineering schemes to improve the speed, which is pre filtering all zero coefficient of DCT blocks and MMX based computing in parallel, are mixed together. These methods can decrease the time of video coding by a wide margin without increase of outputting bit rate and decrease of image quality. Experiment results indicate that the video coder implemented by using these schemes is sufficiently satisfied with applications such as IP videoconference, visual telephone, and so on.