Grants and Contributions:

Grant or Award spanning more than one fiscal year. (2017-2018 to 2018-2019)

Digital video requires a huge volume of data and must be compressed before it can be stored and/orx000D
transmitted. TV broadcasters, video game manufacturers, video content distributors use video compression inx000D
their products and/or services. AMD designs and manufactures graphics cards and microprocessors for tablets,x000D
gaming consoles, embedded devices and cloud servers. These products must yield the best possible videox000D
quality. In video compression, there is an inherent trade-off between bitrate and video quality. Obtaining thex000D
best video quality for a given bitrate is, therefore, a crucial task. The bitrate of a video sequence can be reducedx000D
by reducing its resolution prior to encoding or by using a larger quantization parameter during encoding. Whichx000D
of these two options yields less quality loss depends on the video content as well as the available networkx000D
bandwidth. The goal of this research project is to design a machine learning algorithm to make a run-timex000D
decision of whether to encode a video picture (or group of pictures) at the original high resolution, or to reducex000D
resolution, encode the lower resolution version with a smaller quantization step, decode and upsample atx000D
receiver side with expectation to achieve the best quality for a given bit rate. By utilizing the optimalx000D
resolution/quantization step combination, our developed adaptive video resolution adjustment scheme canx000D
result in significant bitrate savings for a target quality or significant quality improvements for a target bitrate.