Ontario-based SSIMWAVE uses algorithms that mimic the average human or even a studio ‘golden eyes’ executive, perceiving video quality in the way they do. The company claims it is the only measurement technology that achieves better than 90% correlation accuracy between computed objective virtualisation and subjective human opinion scores across all video content, and was at ANGA COM this week to convince European media companies of its benefits.
Monitoring using the SSIMWAVE solution has a range of applications, from helping to choose encoding vendors to optimising the end-to-end content delivery workflow. It can be used to uphold studio expectations for the accurate portrayal of studio intent. Benefits from optimising the workflow include improved quality and reduced cost, and proactively eliminating customer complaints.
The SSIMWAVE technology is designed to replace proxy metrics with what is close to an absolute measure of human QoE, according to executives at the company. It is automated and can be used within an end-to-end monitoring system. Its inventors reckon its ‘psychovisualisation’ approach gives a much better idea of the real QoE when compared to other video assurance technologies.
Zhou Wang, Chief Science Officer at SSIMWAVE, has spent 20 years researching the video qualitative experience and how to assess it, and in 2004 he developed the Structural Similarity Index Method or SSIM, an algorithm that was so effective at mimicking human perception that Wang was awarded the 67th Engineering Emmy Award for co-inventing structural similarity. Since then an improved metric called SSIMPLUS has been introduced.
The first use of this technology was to predict what the average viewer would think about the quality of video, using a range of measures from group studies. SSIMWAVE refers to this as the ‘naïve viewer’ mode. More recently, an ‘expert viewer’ mode has been added. This predicts what a studio ‘golden eyes’ viewer would think of video and acknowledges their much higher expectations, including detail around colour.
The naïve viewer mode is suitable for platform operators who care about end-user satisfaction. The expert viewer mode will allow studios to judge distributors. In both cases, any failure to match expectations can be sourced back to key stages in the content delivery cycle where quality degradation occurred.
The expert viewer mode stems from a project with the American Society of Cinematographers, who are naturally keen that video processing and delivery does not trash the creative intent of the directors and producers. According to Wang, studio creatives are sometimes so appalled by the product that reaches consumers that they never watch television at home. “Distributors do their best but there was never a way to measure their output, and nothing like an SLA (Service Level Agreement) with a studio.”
Wang says studios and production companies hope they can introduce more control over what the delivered content looks like. “If you are hitting a 65% quality measure [against the automated QoE testing that uses the expert viewer algorithm] then perhaps you can ask that the figure is improved to 70% by next year and 75% the year after that. Or maybe the scores start to influence which distributors you work with.”
The SSIMWAVE software can be installed before and after every major process in the distribution workflow, namely encoding, transcoding, packaging, origin serving, CDN distribution and reception on end-devices. If the technology is as good as the company claims, this is the automated equivalent of having a consumer test group sat on each side of the transcoder or packager, giving their scores on the content as it goes in and comes out. In the case of the expert viewer mode, it is like having a studio ‘golden eyes’ at every point in the delivery chain.
The SSIMPLUS solution is already deployed with cable operators in North America, where it is used to monitor over 5,000 TV channels with a software presence next to all the key delivery waypoints (i.e. encoding, transcoding, packaging, origin, CDN and CPE). Over 30 million subscribers are covered by the networks where it is used.
One company using SSIMWAVE – yet to be announced officially – is Dolby, which is using the technology to validate whether its Dolby Vision high dynamic range (HDR) format is being used properly, with a particular interest in Wide Colour Gamut outputs.
The rest of the video monitoring market has been using proxy measures to try to understand viewer perception of quality, according to Carlos Hernandez, Vice President Sales at SSIMWAVE. “With this technology, we can understand and predict what the customer experience is really going to be when they receive the video. This solution transforms the way video quality is measured but it also results in activations – it means you can improve all the processes used to distribute content.”
The SSIMPLUS solution can be used to enhance encoding, for example. A ‘SSIMPLUS saliency map’ shows the parts of an image that humans are focused on, which might be a face, the logo on a carton of apple juice, and the profile of a gun in a drama scene. That can help guide the way you encode a frame or group of frames.
There are other ways the SSIMPlus technology can be used to optimise video delivery, according to the company that invented it. If you discover that viewers – whose opinions are effectively represented via the automated analysis – cannot notice a difference between two ABR bit-rate offerings, one of them can be removed, thus reducing the size of the ABR ‘ladder’ and saving on processing and storage costs.
SSIMWAVE was at ANGA COM this week, having decided to target European media owners and platform operators more actively. The company does have some (unnamed) customers in the UK, although not for end-to-end monitoring. In these instances, the human perception analysis is being used to help choose vendors, including in shoot-outs but also by figuring out the best configuration of different vendors at different stages of the delivery chain.
According to Wang, optimising workflow technology choices is an important application for SSIMPlus. You could first use it decide what codec to use and then the precise implementation of that codec. Then you could look for the encoder vendor that best matches these choices, and subsequently source the statistical multiplexer that works best with that encoder.
Hernandez says the business case for SSIMPlus video analysis can be made based on operational cost savings – it does not even rely on the likelihood that higher video quality helps retain subscribers. Cost savings can be made on bandwidth – like when dropping superfluous ABR profiles. Checking encoder and transcoder claims – which is another use case – can ensure promised bit rate efficiencies are indeed delivered.
Wang adds that QoE issues can be surfaced early, thanks to human-standard perception testing in the content delivery chain. Proactive remedies can be sought. This results in fewer costly, quality-related customer calls.
