Datenbestand vom 21. April 2021
Tel: 089 / 66060798
Mo - Fr, 9 - 12 Uhr
Fax: 089 / 66060799
aktualisiert am 21. April 2021
978-3-8439-3461-9, Reihe Ingenieurwissenschaften
Energy Efficient Video Decoding
245 Seiten, Dissertation Universität Erlangen-Nürnberg (2017), Softcover, A5
In the past years, video streaming services for mobile applications have become a popular application for consumers all over the world. Streaming services like YouTube or Netflix provide the possibility to watch movies, short films, or television (TV) programs anywhere and anytime. As mobile devices depend on batteries for energy supply, it is highly desirable to have energy efficient software and hardware such that the operating time is as high as possible. In this work, the energy needed for decoding a given coded video is addressed in detail. Measurements indicate that during streaming and playback, the major part of the energy the device consumes is needed for the decoding process if a software decoder is used. Hence, research aiming at reducing this energy is a worthwhile task.
In this respect, three major topics are addressed in this work. The first topic deals with the energy analysis of practical decoding systems. Therefore, a measurement setup is constructed and discussed which is able to perform real energy measurements for various decoders and codecs. The results are used to compare the energy efficiency such that optimal decoders can be found in state-of-the-art decoder devices and implementations.
The second topic addresses the modeling of this decoding energy. Therefore, the coded videos are analyzed for properties like resolution, bitrate, and employed compression tools. Then, the relation between these properties and the decoding energy is evaluated. As a result, it is shown that these properties can be used to accurately estimate the complete decoding energy of a given coded bit stream. In a practical application, these estimations can be used to give an early feedback on the remaining operating time of the device. Furthermore, these models can be used to determine the energy consumption that is related to certain compression tools. With this information, the energy efficiency of the tools can be compared.
Finally, the third topic deals with another application for such models. It is shown that they can be used to encode bit streams that consume less energy on the decoder side.
Therefore, an encoder is developed that takes the decoding energy as estimated by the models into account. As a result, it is shown that at the expense of a bitrate increase of 15%, the decoding energy can be reduced by 15% on average. Highest savings are shown to reach 25%.