UBC Theses and Dissertations
Priority progress decoding Kirsh, Lowell
This thesis presents a framework for decoding video in the absence of sufficient computational resources. Most current decoding systems require there to always be sufficient resources available. We have implemented a video decoding module for the QStream video streaming system to adapt the decoding process to the available resources. A distinguishing characteristic of our framework is the ability to decode a video in priority order and drop low priority work if there are not enough resources to fully decode the video. In our particular case we adapt by dropping frames. Additionally, we make no use of a feedback loop as part of the data dropping mechanism. This is important because the variability of video content makes it difficult to predict the relationship between CPU requirements and video content. We argue that due to the inherent variability in both content and devices that an adaptive approach is necessary. We have found our approach to decoding to require minimal computational overhead.
Item Citations and Data