“We want to become synonymous with good user experience for video,” says Abdul Rehman, co-founder and chief executive of SSIMWAVE. The start-up, which was spun out of the University of Waterloo in Canada in 2013, wants to improve how people watch videos online.
The company’s technology is based on a family of algorithms developed by Zhou Wang, a computer engineer at Waterloo and Rehman’s PhD supervisor. The structural similarity, or SSIM, algorithms can predict how someone will perceive the technical quality of a video, and thus help to ensure that viewers have the best possible experience. Software scoring based on the SSIM algorithms reflects the ratings given by real viewers with 95% accuracy, says Rehman.
Getting to that point took a long time and a huge amount of data. Wang and his students — Rehman was the fifth doctoral researcher — spent ten years conducting hundreds of subjective experiments to study the human visual system and how it responds to video content. For example, how different parts of the video capture the viewers’ attention, and how variations in contrast affect how the images are perceived.
The algorithm also takes into account how various technical properties of a video affect the perception of quality. Higher resolution does not mean higher video quality by default, says Rehman. “Jamming more and more pixels in won’t do you any good if you’re not doing a better job at preserving overall quality,” he explains. Resolution, frame rate, and dynamic range — the ratio between the darkest and lightest areas of the picture — all need to be taken into account.
The software also looks at the impact of parameters such as screen size, brightness and pixel resolution on different devices and the viewing conditions. Watching a video on an iPhone is very different from watching it on a television, he says. “If you do not take that into account, then you’re introducing that much error into your quality measurement.”
Rehman compares SSIMWAVE’s knowledge of audience perception to the data accumulated by another giant of the technology industry. “Google understands people’s search patterns,” he says. “We understand their perception of video quality.”
SSIMWAVE’s software helps companies to measure, manage and optimize the viewing experience of videos streamed over the Internet by adjusting the speed of transmission, known as the bit rate, as well as technical aspects of the video such as resolution.
But bitrate does not necessarily mean higher visual quality — using the same bit rate to encode different video content can result in dramatically different visual quality. By adjusting other parameters, SSIMWAVE can achieve the same perceptual quality at a lower bitrate, saving bandwidth and money. “We felt we had to commercialize the software because the industry needed to move from data quality to perceptual quality,” says Rehman.
The company’s clients include the online auction site eBay, the Spanish telecommunications company Telefonica and the computer graphics card manufacturer AMD. SSIMWAVE is also working with the American Society of Cinematographers.
Rehman attributes the company’s success to the university’s support. The university takes no cut of the intellectual property developed by its faculty and graduate students, he says. Researchers own 100% of it. “A university’s ‘product’ is their intellectual property, so for them to say they don’t want it is unique,” he says. “I don’t know of any other university that does that.”
Originally Published on https://www.nature.com/nature/journal/v545/n7654_supp/pdf/545S12a.pdf?origin=ppub