Automatic Video QoE Assessment

SSIMWave’s automatic video QoE assessment technology is at the backbone that enables SSIMWave’s file-based and live-SQM software products to accurately predict end consumers’ visual QoE in real-time on a per network, per user or per service basis.

At the core of the SSIMWave technology is the SSIMplus algorithm, which is an automatic scoring method that predicts how an average human evaluates video quality. It is designed to replace human visual assessment in the application scenarios where visual inspection is slow, expensive and unstable. SSIMplus has a unique package of features unprecedented by any existing method in the academic literature or in commercialized products. These features include:

  • High accuracy
    SSIMplus predicts video QoE of an average observer more accurately than any other algorithm, including the award-winning SSIM index.
  • Straightforward scoring
    SSIMplus provides straightforward predictions on what an average consumer would say about the quality of video on a scale of 0-100, evenly spaced to five categorizes of bad, poor, fair, good, and excellent, respectively.
  • High speed
    SSIMplus can be computed in real-time with purely software implementation up to Ultra HD video.
  • High precision
    SSIMplus predicts video quality up to pixel level, allowing the flexibility to score video quality on a per-segment, per-frame, or even per-pixel basis.
  • Cross-device assessment
    SSIMplus predicts how an average observer experiences different visual QoE when viewing the video on different devices (smart phones, tablet, TV, etc.).
  • Cross-resolution assessment
    SSIMplus compares videos of different resolutions and provides corresponding predictions of visual QoE.
  • Cross-content assessment
    SSIMplus provides meaningful unified scoring for video content with large variations in complexity and the type of content.

SSIMplus is the outcome of more than ten years of advanced research after SSIM. SSIMplus inherits some of the fundamental principles of SSIM but is far beyond SSIM. For example, SSIMplus incorporates vision models to take into account the impact of viewing device and viewing conditions, such that the variations in visual QoE when watching the same video on different viewing devices with different viewing conditions can be well predicted. SSIMplus uses attention models to significantly improve the quality prediction performance. SSIMplus takes into account both spatial and temporal variations of video signals in assessing their quality. SSIMplus preprocesses videos so as to assess videos with different spatial resolutions or videos that are misaligned temporarily. SSIMplus also incorporates state-of-the-art models to assess the quality of high dynamic range (HDR) videos. In addition, SSIMplus combines advanced modeling with smart engineering implementations, leading to an elegant solution best matches the practical needs of the video delivery industry.