Algorithms for video compression

本研究计划的目标是通过以下方式来推进压缩技术：1)研究从根本上不同的帧预测方法；2)研究选择编码参数以提高视觉质量的方法。第一个目标将通过识别在每个视频帧中可以在其他帧中找到的场景特征来实现，尽管视点，比例或照明存在差异。通过匹配这些特征，我们将建立帧之间的对应关系，从而提高压缩比。本提案的第二个目标将通过集成视频编码器来实现，这意味着客观地评估视频的视觉质量，并调整编码器参数以最大限度地提高视觉质量。这两个目标都需要计算复杂的算法。然而，对于离线压缩，并且考虑到处理器计算能力的持续增长，这将不会成为压缩技术进步的障碍。

The objective of this research program is to advance compression technology by 1) investigating fundamentally different approaches to frame prediction, and by 2) investigating means to choose encoding parameters to increase visual quality. The first objective will be achieved by identifying in each video frame scene features that can be found in other frames despite differences in viewpoint, scale or illumination. By matching these features, we will then establish a correspondance between frames that will improve compression ratios. The second objective of this proposal will be achieved by integrating, into a video encoder, means to objectively evaluate the visual quality of video, and adjust encoder parameters to maximise the visual quality. Both objectives will require computationally complex algorithms. However, for offline compression, and taking into account the continuing increase in processor computational power, this will not be an impediement to the advancement of compression technology.