Compression Coffee Break: AI for Content Adaptive Encoding: Get the Big (and Better) Picture, Literally
When you’re watching a car chase in a movie, how important are the trees or the mailboxes or the telephone poles? Unless one of the cars has found a way to travel atop the telephone lines, not as important as the cars and the drivers themselves. And how important is being able to watch that car chase in rich vibrant colors and at top speed, the way it was meant to be seen, regardless of the millions of other viewers who might be watching, leaning hard on the bandwidth resources?
At Synamedia, we know how important it is to be able to see any type of content in the highest quality while using the lowest number of bits. That is why we have developed new codecs that identify and compress the media content and automatically adjust bitrate and target quality level.
Synamedia’s ability to deliver stunning picture quality with automatically adjusted encoding is significant particularly now, during a time when, due to stay-at-home measures from the COVID-19 pandemic, the world has seen a surge in Internet traffic. The problem is, video streaming services have limitations with regards to scale. Some services have dealt with the network congestion by reducing streaming bandwidth. Netflix, for example, reduced bit rates across streams in Europe to cut traffic by 25%. It may take care of the bandwidth problem, but ultimately the end user’s experience is not the same. You will still be able to access the video content, but at a lower quality.
AI for Content Adaptive Encoding
Enter machine learning (ML) and artificial intelligence (AI) for content adaptive encoding (CAE), which allow viewers to watch car chases, soccer matches and the local news with equal fidelity. With new broadcasting standards, such as ATSC 3.0, supporting 4K and ultimately 8K streams over-the-air, we’ve been laser-focused on developing new compression techniques and advanced encoding technologies to deliver stunning low-latency live experiences while optimizing bandwidth.
AI and ML will allow us to not only compress whole or partial frames individually, but to also predict future action based on past ones, determine the areas within each frame that will matter least to viewers and compress those areas more than the others. By leveraging AI and ML as a tool for codec optimization, we can match the high sensitivity to image quality and abrupt changes that are detected by the human eye. This helps us take a substantial step as we aim to deliver top-quality live streams by significantly improving adaptive bit rate encoding at the live program or event level.
Benefits of Automated CAE
Whether the content features a high-speed sports game, a news broadcast with little movement or a wildlife documentary with vibrant color, our team of R&D compression specialists is using neural networks to get even more granular and recognize different scenes and logos. This allows the encoding to be automatically adjusted based on specific program segments to minimize the bits used while maintaining a consistently high video quality. Not to mention the streaming and storage cost savings – we can deliver consistent video quality with bandwidth savings of 30-50 % compared to constant bitrate.
To learn more about how we offer bandwidth savings while maintaining consistent video quality, check out our white paper on Smart Rate Control.
This blog is part of our series of Compression Coffee Break posts – aimed to provide you with small “snacks” of information about the benefits of compression and other video distribution, delivery and processing, along with the industry challenges they will help overcome. Keep an eye out on our blog for more – we promise delays will be minimal, just like watching live sports events powered by our technology.
About the Author
Bart is responsible for marketing Synamedia’s video network product and architecture portfolio, covering cloud, workflow automation, ABR technology and microservices. With over 20 years’ experience in the video industry, Bart demonstrates how Synamedia’s solutions and services help service providers to deliver premium video services securely and reliably to any screen.
Prior to joining Synamedia, Bart spent 12 years with Cisco, and before that was at Scientific-Atlanta, which was subsequently acquired by Cisco. Before that he held various engineering and product management roles.
Bart holds a Master of Science in Applied Engineering from the KaHo Sint-Lieven, currently University of Leuven, in Belgium.