Why do video sources fall short of expected quality?

There is this prevailing myth in the video industry that source content is more or less OK, but the truth is that video has a long journey to the subscriber from the camera at the video shoot or the stadium truck. From the final director’s cut through mezzanine files it has already passed through several steps of compression before it reaches your encoders. And a lot can happen in these steps:

• Banding and other artifacts can be introduced as part of reprocessing or preparing the mezzanine files;
• Overcompression can lead to distorted colours, skin tone deviations;
• De-noising makes content more susceptible to banding;
• And sometimes there is the outright upsampling of content so that it fulfills contractual agreements about minimum bitrate, but doesn’t correct bad video quality.

Our experience in processing more than 10 million hours of video each month shows that setting your Source Target quality at > 80 SSIMPLUS Viewer Score will guarantee good viewer experience levels at your subscribers’ devices, if no other issues and artifacts are introduced with downward processing.

For premium content and quality tiers, strive for a SSIMPLUS Viewer Score at the source of >90.

How can you stop bad video in its tracks to keep subscribers happy and prevent churn? To help you out SSIMWAVE developed the ultimate Gatekeeper — SSIMPLUS single-ended perceptual quality metric based on more than 30 years of key research into the Human Visual system.

Achieve FR accuracy using SSIMPLUS no-reference approach

SSIMWAVE’s solution captures spatial, temporal and joint video impairments by inspecting assets down to their constituent pixels. With a higher than 90% correlation with the most accurate full-reference metric (SSIMPLUS), SSIMPLUS No-Reference QoE measures Viewer Experience in real-time.

It is the only single-ended quality assessment that can provide meaningful scores across all resolutions, frame rates, dynamic ranges adapted to target display devices.

What is unique about SSIMPLUS no-reference metric is that it doesn’t use bitrate and encoding configurations — as these can be easily fooled. For example, some content providers will take a 5Mbps source and upsample it to 35Mbps to fulfill contractual agreements, but the video quality will be bad. SSIMPLUS cannot be fooled by the higher bitrates unlike other traditional algorithms because it looks at the pixels. Our deep-learning based algorithm (unlike all other ML/AI approaches) learns directly from the pixels as it is trained with an extensive library of complex content.

Content validation use cases that do not have a mezzanine source available can fully rely on SSIMPLUS single-ended metric to validate the Viewer Experience at any point in a delivery chain. Typical deployment scenarios include:

• Source validation — how would OTT services, D2Cs, MVPDs that receive hundreds if not thousands of videos from different providers evaluate the quality of the assets? How to decide if those videos are worth processing further?
• Encoder validation — in other situations, content delivery teams might have access only to the encoder output without the source file. How to ensure that these encoded outputs are of the desired video quality?
• Customer premises validation — another very common situation is when providers have access only to the HDMI out of a user device. Similarly to the other use cases, they need a Viewer Experience metric they can rely on to measure and understand if and how the workflow processes and vendors have affected the Viewer Experience and what is the degradation vs. the targeted SSIMPLUS score.

In all of the above situations, SSIMPLUS single-ended quality assessment provides the needed KPI to make an informed and accurate decision. Request your free trial today