3 ms·
That's curious. My observation has been that most of the streaming services have really high compression ratios with 4k streams. With videos like this one, whic
by cbsmith 6y ago
That's curious. My observation has been that most of the streaming services have really high compression ratios with 4k streams. With videos like this one, which are built from still images, I imagine you still get significant improvements in quality, but for content that isn't so easily compressible (e.g. action movies), I've seen significant effects from compression in 4k streams as compared to "regular" 4k source material.
As far as "wasting computing power"... I hear some people play video games while watching movies... on the same machine. So that's probably what that is about.
- crazygringo 6y agoThat's exactly what I'm saying, though. Because YouTube (like other streaming services) aggressively compresses both 1080p and 4K, but the 4K version still has a much higher bitrate, watching the 4K stream on a 1080p monitor will be significantly better quality than watching a 1080p stream on a 1080p monitor. I'm not saying you won't see any compression artifacts at all, of course if it's an scene you will. But the 4K version is still noticeably way better than the 1080p one, even in an action scene. It's a higher bitrate so there's more detail no matter what.
- cbsmith 6y agoYeah, what I'm saying is that because the 4k streams are trying to convey more entropy than the 1080p streams, I've observed that even though their bitrate is higher, the resultant compression effects are far more severe, to the point where the overall experience is better with 1080p, sometimes even if you have access to the full 4k resolution. I've seen similar effects with audio content (after a certain point, increasing sampling rate significantly more than bitrate has a deleterious effect). From an information theory perspective, you are obviously correct that there is more information in the 4k streams and therefore you will get more detail from them. My observation however has often been the opposite. My attempt to rationalize leans towards notions that the lossy compression techniques we use with video codecs still have significant impedance with the "compression" in our neuro-opthalmologic system.
- crazygringo 6y agoOh now I understand what you're saying -- that makes sense. It's not what I've observed, but I definitely can imagine how settings for the compression, or what the algorithm concentrates on, could make that happen. I'll have to pay close attention now! And your rationalization probably makes sense -- I read the h.266 codec is supposedly focusing specifically on perception improvements in 4K/8K, so it may be addressing what your'e perceiving specifically. Funnily enough, this reminds me of what I did with JPEGs on websites when Retina first became a thing. Instead of serving up different assets depending on resolution, I discovered that serving up a 2x resolution JPEG to everyone, with really crappy quality, was superior to serving up 1x with high quality. For 1x screens, blockiness was inherently shrunken because of JPEG's fixed block sizes, and for 2x screens the low quality was harder to perceive because it was at a small scale anyways. How compression interacts with scale/resolution is not a necessarily intuitive thing.
- cbsmith 6y agoThat's a brilliant conclusion on the JPEG hack, and yes, that's precisely what I've observed. You really notice it if you compare to even 1080i with high bitrates (OTA HD TV for example). It's weird.