Save your FREE seat for Streaming Media Connect this August. Register Now!

What Are Objective Quality Metrics?

Watch the complete panel from Streaming Media East Connect, How to Fine-Tune Your Encoding with Objective Quality Metrics on the Streaming Media YouTube channel.

Learn more about streaming video quality metrics at Streaming Media West 2020.

Read the complete transcript of this clip:

Jan Ozer: Let's start with what objective quality metrics are. Essentially, they are mathematical formulas that attempt to predict how human eyes would rate videos. So they have no value of their own. Really, it's just, how well and how accurately do they predict how humans will view the videos.

Some examples that you probably have heard of include Mean Opinion Scores--one of the first--PSNR (Peak Signal-to-Noise Ratio) and SSIM (Structural Similarity Index) are some early ones, SSIMPLUS, and VMAF, which is the metric invented by Netflix.

Over time, these metrics have become more accurate, because they've incorporated a couple of things. Number one, the initial metrics were just math. They measured the difference between the compressed video and the source video, and came up with a measure of that difference.

And that's kind of the Mean Square Error in PSNR. SSIM was the first one to try and gauge how much humans actually cared about those differences. Some differences you notice more, some differences you notice less, and that's why most people who use metrics consider SSIM superior to PSNR. And then once you get to the more recent metrics--SSIMPLUS and VMAF--these incorporate math in their basic algorithms, plus the concept of how humans perceive the differences and also machine learning elements, particularly for VMAF.

As we go along, these metrics get better and better, although no metric is 100% accurate. You can always find use cases where they don't work. But at some point, they are going to be able to mimic the human visual system entirely.

Streaming Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues
Related Articles

Review: Moscow State University Video Quality Measurement Tool 14.1

Unlimited file analysis and Python interface highlight the new features

How to Balance Encoding Time and Quality

Jan Ozer explains how to choose presets and tweak encoding parameters when prioritizing speed or quality, and how to find the perfect balance.

How to Find the Optimal Data Rate for Encoding

Jan Ozer explains how to use constant rate factor encoding and per-title encoding to find the optimal data rate for encoding different types of video files, ranging from talking heads to high-action movies.

Applying Objective Quality Metrics in Your Own Encoding

Jan Ozer discusses real-world applications for per-title and per-category encoding, where applying objective quality metrics to recalibrate an encoding ladder is effective and also times when it's not as helpful.