Increasing the video quality of your streaming service can have noticeable impact on overall user engagement. Customers don’t just want fast startup times; they are very sensitive to video quality issues like blurring, blockiness, color distortions – or generally too low bitrates. Large-scale studies have shown that improving video quality also improves the time users spend with a service.
We provide you with the right tools to measure and improve video quality – and make your customers happier. You can perform these measurements while ingesting video, during transcoding, or at the client side.
With the proper video quality testing solution, you will be able to catch delivery problems early on, optimize the entire transmission chain, and prevent user churn.
Video quality is more than bitrate and resolution. Knowing that a video is “HD” or 4K is not enough. It is important to measure video quality as your customers experience it.
Video quality models are the algorithms that are used to determine the visual experience that a video delivers.
There are many such models on the market, but not every model fits every use case. From our year-long experience in video quality standardization – and working on real-life cases with the industry–, we know which model is suitable for which purpose. Depending on the application area, we recommend to use different models for:
Our video quality testing solutions are as flexible as your use-case. We will pick the right models for your application.
The best video quality model is not usable in practice if it takes too long to compute. As a provider, you want fast processing and low resource usage, while getting the most accurate measurements.
Pixel-based models like PSNR, SSIM (and its variants), or VMAF work by inspecting and comparing each pixel of every frame of the source video and transmitted video. This takes enourmous processing resources and is not scalable, particularly when the content is not known before (e.g. live TV).
By focusing on so-called parametric, bitstream-based models we can leverage metadata in the video stream to estimate the video quality. This means that the video doesn’t have to be decoded, which in turn saves precious processing time and storage space, and therefore significantly reduces overall costs. The quality measurements can be scaled virtually without limits, to hundreds of thousands of streams. With low-complexity models, you can for example distribute the quality calculation in the cloud and effortlessly scale your computing resources according to the demand.
All video quality models give you an estimation of the quality – for example between 0 and 100, or on a scale from 1 to 5. However, interpreting this scale and comparing the numbers is not always possible. Standards bodies like the ITU (International Telecommunication Union) have formally defined the Mean Opinion Score (MOS) scale – and algorithms that can be used to measure video quality in a valid way.
Standardization plays an important role in creating video quality measurement algorithms that have been extensively validated by a large consortia, using thousands of human viewer’s ratings. They are among the most accurate and reliable tools that operators can use.
We implement our quality measurement solutions based on the world’s first and only HTTP Adaptive Streaming QoE Model, recommended by the ITU-T: the ITU-T P.1203 family. It has been the product of several years of joint research and has been shown to significantly outperform other models in terms of accuracy and speed.
On top of the existing standards, we cooperate with our academic partner, TU Ilmenau, to extend and validate the video quality models so that they are fit for next-generation codec technology.
ITU-T P.1203 can use different information to predict video quality.
Sign up and find out what’s happening in the world of video streaming.