While the energy consumption of many aspects of video streaming—such as encoding and decoding—has been thoroughly analyzed and optimized, the energy consumption associated with video quality analysis is often overlooked. Typically, video quality analysis is performed using objective metrics that attempt to simulate human perception of quality.
Metrics for measuring video quality can vary widely, ranging from simple pixel comparisons to complex neural networks. Because of these very different approaches, not only does accuracy vary, but energy consumption and complexity also differ significantly among the various metrics.
The goal of this thesis is to investigate various video quality metrics in terms of energy consumption and accuracy. Depending on the results of the energy measurements, the aim is then to identify ways to optimize the video quality analysis in terms of energy consumption without significantly compromising accuracy.
Prerequisites
- Solid background in image and video compression
- Experience in Python
- Interest in video compression and video quality