Joint Project For Digital Mobile Multimedia Communiucation Research

在去年活动成果的基础上，本年度主要开展了三方面的研究工作：关键技术的比较与评估、验证模型的开发、基于视频文档体系结构模型的视频编码方法的开发与实现。在根据可能的算法的功能、测试序列的类别及其比特率对可能的算法进行分类后执行测试。采用MPEG1作为CIF格式的锚点，H：263作为CIF的QCIF锚点。结果表明，锚点和基于块的方法在10kbps左右的较高码率下表现出更好的性能，并且帧速率是影响解码图像的最大因素之一。有前景的关键技术候选包括长短记忆结合、帧内小波编码、全局MC和局部仿射MC的结合、块划分、匹配跟踪等。263扩展运动矢量和高级预测模式、二进制Alpha的四叉树编码和灰度Alpha的四叉树编码加VQ、填充边界编码等技术。最后，基于视频文档架构的编码方法的开发和实现。提出了一种基于结构化视频概念的编码格式。根据这一概念，将一组图片信息分解为若干结构要素，保存为素材作为视频序列的一部分。在视频编辑中，这些材料被有选择地处理或合成以生成一个全新的序列。编辑模拟表明，编辑过程的工作量得到了显著改善。

Based on the last year's activity results, three major research activities have been done during this year : Comparison and evaluation of key technologies, Development of verification model, Development and implementation of video coding method based on a structuring model called Video Document Architecture.First, on comparison and evaluation ofkey technologies. Test were performed after classifying the possible algorithms by their functionalities, the class of test sequences and their bitrates. MPEG-1 was used as an anchor for CIF format, and H : 263 for QCIF of CIF.Results show that the anchor and block-based methods tend to behave better performance at higher bitrates around 10kbps, and that the framerate is one of the most influential factors on the decoded images. Candidates of promising key technologies include combination of long and short term memories, wavelet coding on intra frames, combination of global MC and local affine MC,block partitioning, matching pursuits, etc.Next, on development of verification model. Long lasting discussion finally reached an agreement that the verification model should have the plural input video contents as video object planes, and that it should hold the following technologies : H.263 extended motion vectors and advanced prediction modes, quadtree coding for binary alpha and quadtree coding plus VQ for grayscale alpha, boundary coding by padding.Finally, on development and implementation of coding method based on Video Document Architecture. A coding format based on the concept of structured video is developed. According to this concept, a set of picture information is separated into some structural factors, saved as materials to be part of video sequence. When in video editing, these materials are selectively manipulated or synthesized to generate a whole new sequence. The editing simulation indicate that the workload for editing process becomes significantly improved.