API Build-data JSON Resources
Theme

Choose how MSRBot.io looks on this device.

Auto (match system) Light mode Dark mode

Preference is stored in this browser only.

SMPTE Motion Imaging Journal ( Volume: 131, Issue: 3, April 2022)
[ACTIVE]

Machine Learning for Per-Title Encoding

Metadata

Publisher
SMPTE — White Plains, NY, USA
Doc Type
Journal Article
Content Type
Original Research
Abbreviated Title
SMPTE Mot. Imag. J
Volume
131, No. 3, pp. 42–50
Abstract
Video streaming content varies in terms of complexity and requires title-specific encoding settings to achieve a certain visual quality. Classic “one-size-fits-all” encoding ladders ignore video-specific characteristics and apply the same encoding settings across all video files. In the worst-case scenario, this approach can lead to quality impairments, encoding artifacts, or unnecessarily large media files. A per-title encoding solution has the potential to significantly decrease the storage and delivery costs of video streams while improving the perceptual quality of the video. Conventional per-title encoding solutions typically require a large number of test encodes, resulting in high computational times and costs. In this article, we describe a solution that implements the conventional per-title encoding approach and uses its resulting data for machine learning-based improvements. By applying supervised, multivariate regression algorithms like random forest regression, multilayer perceptron (MLP), and support vector regression, we can predict video quality metric (VMAF) values. These video quality metric values are the foundation for deriving the optimal encoding ladder. As a result, the test encodes are eliminated while preserving the benefits of conventional per-title encoding.
Publication Date
2022-04-01
DOI
10.5594/JMI.2022.3154836
ISSN
Print: 1545-0279 | Electronic: 2160-2492
Link
https://doi.org/10.5594/JMI.2022.3154836
Author(s)
Daniel Silhavy
bio
Daniel Silhavy is a scientist and project manager at the Business Unit Future Applications and Media (FAME) of the Fraunhofer Institute for Open Communication Systems (FOKUS) (Fraunhofer FOKUS, FAME), Berlin, Germany. He is also the lead developer of the dash.js project and the development coordinator for the 5G-Media Action Group (5G-MAG) Reference Tools. His main area of expertise includes adaptive bitrate (ABR) streaming focusing on MPEG-DASH, digital rights management (DRM), 5G media streaming, and video-encoding. He actively contributes to standardization work done in DASH-IF, 5G-MAG, and Consumer Technology Association-Web Application Video System (CTA-WAVE). Silhavy has published various technical papers covering topics around ABR media delivery and streaming analytics.
Christopher Krauss
bio
Christopher Krauss received a PhD degree (Dr.-Ing.) from the Technical University of Berlin (TU Berlin) with his dissertation “Time-Dependent Recommender Systems for the Prediction of Appropriate Learning Objects.” He works as media & data science lead at Fraunhofer FOKUS, FAME. He specializes in the research and development of topics dealing with machine learning, data science, technology enhanced learning, and connected TVs, and has been involved in multiple public funded projects (e.g., mEDUator, TripleAdapt, EXPAND+ER WB3, Smart Learning I & II, Digi-Hand, FI-Content I & II, User Centric Networking, and Global ITV) and managed many projects for different national and international industry customers.
Anita Chen
bio
Anita Chen has over ten years of technical management in the U.S., Germany, and China, with a focus on the research and development, automotive, health, and media industries. Chen received a BS degree from Business Information Technology, Virginia Tech, Blacksburg, VA, in 2010, and an MS degree from Information Technology Management, TU Berlin, in 2020.
Anh-Tu Nguyen
bio
Anh-Tu Nguyen is a researcher and software developer at Fraunhofer FOKUS, FAME. He received bachelor’s and master’s degrees in industrial engineering from TU Berlin. Currently, he is involved in the deep encode project that optimizes video streaming processes using machine learning techniques. He is interested in the applications of machine learning in the computer vision field.
Christoph Müller
bio
Christoph Müller studied computer science at TU Berlin. He received a degree from FOKUS, with the completion of his thesis “Machine Learning-Supported Adaptive Streaming Analytics” in 2018. He is currently employed as a project manager at FAME, where he is the lead technical project manager for “Deep Encode,” an AI-based solution for per-title and per-scene video encoding. His area of expertise lies in the research and development of modern web applications and machine learning in the context of video encoding, adaptive media streaming and streaming analytics, and has published numerous technical papers on these topics.
Stefan Arbanowski
bio
Stefan Arbanowski is director of Fraunhofer FOKUS, FAME. Arbanowski received PhD and MSc degrees in computer science from TU Berlin. Currently, he is managing Fraunhofer FOKUS’ NewTV activities, bundling expertise in the areas of interactive applications, web technologies, streaming and connected TV-channeling those toward networked media environments featuring live, on demand, context-aware, and personalized interactive content.
Stephan Steglich
bio
Stephan Steglich is director of the business unit at Fraunhofer, FOKUS, FAME. Steglich received MSc and PhD degrees in computer science from TU Berlin in 1998 and 2003, respectively. He is managing international and national level research activities and has been an organizer and a member of program committees of several international conferences. He is actively involved in standardization activities in these research areas and gives lectures at TU Berlin.
Louay Bassbouss
bio
Louay Bassbouss is a scientist and senior project manager research and development at Fraunhofer, FOKUS, FAME. He works on future web applications, multiscreen technologies and standards, and 360° video technologies. Bassbouss actively participates in various standardization groups in the World Wide Web Consortium (W3C), Hybrid broadcast broadband TV (HbbTV), and CTA. He is the co-chair of the W3C Second Screen Working and Community Groups and actively contributes to various testing activities in HbbTV and CTA-WAVE.
Keyword(s)
Adaptive Bitrate Streaming, Machine Learning, Per-Title Encoding, VMAF
Copyright
© 2022 SMPTE

Bibliographic Reference(s)

Source Data (JSON)

Full registry record with provenance metadata. Open directly: /api/doc/10.5594-JMI.2022.3154836.json

Reference Tree

Explore all references and references to this document, as a navigable tree.

Open Reference Tree
Reference this Doc

Plain text (ISO 690 compliant)

Preview:
Daniel Silhavy, Christopher Krauss, Anita Chen, Anh-Tu Nguyen, Christoph Müller, Stefan Arbanowski, Stephan Steglich, and Louay Bassbouss; Machine Learning for Per-Title Encoding, SMPTE Motion Imaging Journal ( Volume: 131, Issue: 3, April 2022); SMPTE, 2022. Available at https://doi.org/10.5594/JMI.2022.3154836
Snippet:
Daniel Silhavy, Christopher Krauss, Anita Chen, Anh-Tu Nguyen, Christoph Müller, Stefan Arbanowski, Stephan Steglich, and Louay Bassbouss; Machine Learning for Per-Title Encoding, SMPTE Motion Imaging Journal ( Volume: 131, Issue: 3, April 2022); SMPTE, 2022. Available at https://doi.org/10.5594/JMI.2022.3154836

HTML (ISO 690 compliant)

Preview:
Daniel Silhavy, Christopher Krauss, Anita Chen, Anh-Tu Nguyen, Christoph Müller, Stefan Arbanowski, Stephan Steglich, and Louay Bassbouss; Machine Learning for Per-Title Encoding, SMPTE Motion Imaging Journal ( Volume: 131, Issue: 3, April 2022); SMPTE, 2022. Available at https://doi.org/10.5594/JMI.2022.3154836
Snippet:
<span class="citation">Daniel Silhavy, Christopher Krauss, Anita Chen, Anh-Tu Nguyen, Christoph Müller, Stefan Arbanowski, Stephan Steglich, and Louay Bassbouss; <cite>Machine Learning for Per-Title Encoding</cite>, SMPTE Motion Imaging Journal ( Volume: 131, Issue: 3, April 2022); SMPTE, 2022. Available at <a href="https://doi.org/10.5594/JMI.2022.3154836" target="_blank" rel="noopener">https://doi.org/10.5594/JMI.2022.3154836</a></span>

SMPTE Icon SMPTE's HTML Pub

Preview:
Daniel Silhavy, Christopher Krauss, Anita Chen, Anh-Tu Nguyen, Christoph Müller, Stefan Arbanowski, Stephan Steglich, and Louay Bassbouss; Machine Learning for Per-Title Encoding, SMPTE Motion Imaging Journal ( Volume: 131, Issue: 3, April 2022); SMPTE, 2022
doi: 10.5594/JMI.2022.3154836
url: https://doi.org/10.5594/JMI.2022.3154836
Snippet:
<li>
Daniel Silhavy, Christopher Krauss, Anita Chen, Anh-Tu Nguyen, Christoph Müller, Stefan Arbanowski, Stephan Steglich, and Louay Bassbouss; <cite id="bib-10-5594-jmi-2022-3154836">Machine Learning for Per-Title Encoding</cite>, SMPTE Motion Imaging Journal ( Volume: 131, Issue: 3, April 2022); SMPTE, 2022
<span class="doi">10.5594/JMI.2022.3154836</span>
</li>

Referenced By