By Tony Jones, Principal Technologist, MediaKind
Delivering video content is costly. Networks have limited capacity, content delivery networks (CDNs) charge per delivered bit, and Cloud DVR storage costs can be high, particularly for private copy. The media industry has traditionally concentrated its encoding efforts around Constant Bitrate (CBR) delivery for encoding each representation of Adaptive Bitrate (ABR) video for over-the-top (OTT) delivery services. CBR encoding enables a stable and constant bitrate. When we set a CBR rate, the reality is that we’re using bitrate as a proxy for quality – i.e., it’s a quality level that we’re really trying to achieve, and setting bitrate is a means to get (at least) that quality.
This has been a reasonable approach for traditional broadcast systems using fixed bandwidth connections, but its lack of dynamism is also its most significant drawback. CBR cannot respond to the variance of content complexity, which typically alters scene-to-scene. Too many bits are wasted on ‘easy’ content (e.g., static scenes which provide very little detail), encoding to a bitrate higher than is necessary to achieve the desired quality. At the other end of the spectrum, it can often be the case that ‘complex’ content (e.g., fast-action sports) might not have quite enough bitrate, resulting in possible visible compression artifacts.
So, what’s the solution? Enter MediaKind’s Constant Video Quality (CVQ) encoding technology, a compelling and unique alternative to traditional CBR approaches for processing ABR video. Our in-house video compression experts developed the CVQ model using advanced single channel encoding techniques. CVQ uses advanced data analysis techniques to ensure that OTT video is encoded at a constant video quality rather than a constant bitrate, while respecting target bitrate constraints. In doing so, it avoids the use of unnecessarily high video rates and maximizes the potential compression efficiency of MediaKind encoders – reaching a target and quality that can be met with a lower bitrate. In addition, it can take advantage of the statistical nature to increase instantaneous bitrate during particularly challenging parts of the content to deliver a more consistent quality of experience.
CVQ encoding aims to maintain a consistent, subjective user experience over a wide range of content of varying complexity. As the scenes vary, our sophisticated CVQ model can instantly adjust the bitrate within the encoder according to the spatio-temporal complexity of the content. This achieves a video quality as constant as possible under those constraints. The bandwidth saved through CVQ is then made available to keep a similar quality when more complex content comes along. Therefore, CVQ can save bandwidth and storage costs (e.g., a Cloud DVR or CDN).
The demo video above shows how this works in practice, using advanced encoding techniques that are based on a model of the human visual system. We developed the CVQ algorithm by running multiple encoding experiments over various test sequences at different quantization levels and then visually assessed the encoded streams in terms of comparable subjective quantity. Advanced data analysis techniques were then used to find correlations between quantization levels, spatio-temporal complexity, and subjective quality. This enabled us to build the CVQ encoding model.
Our CVQ technology has been designed to work in conjunction with ABR encoding for OTT delivery services and is set up end-to-end within several MediaKind products, either individually or as part of a unified solution. To date, the CVQ model has been integrated within MediaKind Encoding Live, MediaKind Encoding On-Demand, MediaKind Packaging, and our multi-award-winning VSPP solution.
In our new application paper, “Constant Video Quality: Maximizing video performance,” we provide a detailed analysis that explains how CVQ keeps subjective video quality at a constant level and delivers a more consistent user experience. We also explore how that in conjunction with selective storage for on-demand encoding, it’s possible to dramatically reduce the bandwidth and storage requirements of ABR services and, in doing so, save significant costs attributable to media delivery. If you would like to learn more, please feel free to drop me a message on LinkedIn or Twitter.