TECHNICAL
CAE solutions have since emerged that dramatically reduce the compute requirement while achieving close to optimal results. Jan Ozer has compared various options for H.264 encoding. 6 Additionally innovative AI methods are now being applied to content transcoding to reduce bitrate and improve the visual experience. AI is used to dynamically analyse frames and scenes to tune encoder modes that impact both bitrate and video quality. Bitrate reduction helps lower cost of transmission and cost of storage for media providers. In encoding, parameter choices are impacted by several factors, such as the type of delivery (live, Video on Demand (VoD), and file-to-file (F2F) transfer). Furthermore, media providers use various infrastructures in their data centres for encoding content. These technologies can include specialised software encoders running on CPUs and GPUs with built-in transcoding accelerators. Intel works with the ecosystem that supports these media providers to apply the power of Intel® technologies, such as the Intel® Video Processing Library (Intel® VPL), Intel Xeon Scalable processors, and Intel Data Center GPU Flex Series to their challenges. Intel’s commitment to the ecosystem aims to help lower costs of operations while improving content delivery and thus the viewing customers’ experiences. VisualOn Optimizer Suite VisualOn is a streaming solutions provider that offers a universal, encoder-agnostic, content-adaptive encoding technology. The VisualOn Optimizer suite enables streaming media companies to deliver compelling multimedia content with incredible user experience.
This white paper describes VisualOn Optimizer, its impact on video encoding performance, compares it against other CAE solutions and presents test results showing the benefit of combining the Optimizer with 4th Gen Intel® Xeon® Scalable processors and Intel® Quick Sync Video running on Intel® Data Center GPU Flex Series.
Content Adaptive Encoding
CAE was pioneered by Netflix from 2015 to 2018, with per-title, per-chunk and per-shot encoding. Using their CAE technology, Netflix achieved over 30 percent bitrate reduction 2,3,4 without degrading visual quality, as measured by the Video Multimethod Assessment Fusion (VMAF) score. 5 The approach requires running hundreds of different encodings with different combinations of paraments to select the best results. Such an encoding regimen can be prohibitively expensive and difficult to scale for many companies, especially those with limited budgets or technical capabilities. Building on the strides made by Netflix in encoding strategies, CAE takes video compression a step further by adapting the encoding process to the specific content of each video segment. Unlike traditional encoding methods that apply uniform settings across an entire video, CAE analyses factors such as motion, texture, and complexity within the video to optimise encoding settings dynamically. This results in more efficient compression that preserves quality while reducing file size and bandwidth requirements. These are benefits that are otherwise only achievable through the introduction of new, more complex compression standards that will entail high cost, long time to market, and compatibility issues within the whole ecosystem. Heuristic
Figure 1: VisualOn Optimizer integrates into any streaming workflow.
SEPTEMBER 2024 Volume 46 No.3
Made with FlippingBook - Online magazine maker