1

I noticed when I export my footage at low bitrates (eg. 10 Mbps, 1080p), my 1080p footage looks way blurrier than my downscaled 4K footage.

This is despite the fact that the original footage looks similar in terms of sharpness when viewed at 1080p. Anyone knows what is the problem? This issue disappears at higher bitrates.

Another weird thing is that 10 Mbps 4K downsampled is actually sharper than 50 Mbps native 1080p? I can't get my head around this. I always thought the benefits of downsampling is only visible at high bitrates. Something about the low bitrate seems to affect the 1080p footage but not the 4K footage.

Have a look at the images below - no sharpening applied - tried to get roughly the same framing. Ignore my color grading, I know it's not that great.

Also, I'm using the latest version of Adobe Media Encoder. The 1080p footage in question is 120p if that matters, and the original 1080p footage has quite a low bitrate. However, as seen in the image below, despite the low bitrate, the original still looks relatively sharp.

Original enter image description here

50 Mbps Export enter image description here

10 Mbps Export enter image description here

10 Mbps Export (4K downscaled) enter image description here

1

First let me explain chroma subsampling, as this is the reason why 4k -> 1080p always looks better than native 1080p

Consumer & pro-sumer camera gear, anything not super high end, cuts some corners to save on processing and storage. You'll find this information when you look in your specs and see the 4:2:2 like number. So what's happening is for every 4 pixels, in a square, 2 of them are being used for color (so the top 2 pixels are the same color, and the bottom two pixels are the same color) and all 4 pixels have their own brightness level so that you get the appearance of 4 different colors in this 2x2 block of pixels. This reduces the amount of data by 1/3rd and what's even more popular even in higher end equipment is 4:2:0 which cuts the data in half. The reason it saves so much is because a color requires 3 pieces of information plus 1 piece for brightness, so 4 per pixel, but if you throw away color information but keep half or all the brightness, you can approximate the colors. This is generally imperceivable to the naked eye but it's still a detail shortcut and it's problematic for green screening where you now have multiple mixes of green and non green mixing together and is part of the reason why you see a green halo around people in amateur green screening.

This concept is applied at any level you are recording, so if you record at 1080p your camera is probably doing 422 or 420 and if you kick it up to 4k it's still doing 422/420 but when you down sample 4k to 1080p e, you turn a 2x2 block of pixels into 1 pixel, so now you have all the color and light detail so you are essentially at 444 quality, which is maximum detail.

You can read more about it on Wikipedia: https://en.wikipedia.org/wiki/Chroma_subsampling

  • I understand chroma subsampling, but it doesn't seem to be an issue with the original 1080p footage, unless you're saying that chroma subsampling is more of a problem at low bitrates? – Michael Jun 17 at 9:36
  • What is happening is that the native 1080p footage is subsampled, so the chroma is at 960x540 resolution, but the 4k footage that is scaled down has full resolution chroma. There's an additional factor at play, by using 4 source pixels to create one output pixel, it's a bit like shooting in 10 bit, so there is more information to start with. – stib Jun 20 at 0:08
  • @stib well it's not outright 1/4th the res because the lighting levels are still full 1080p, the color is 1/4th but then the brightness changes the shade of that color as close as the pixel really was, it's a pretty neat optimization and the damage on quality is minimal... tbh it doesn't really explain the guy's problem – TravisO Jul 1 at 20:31
0

Very good question!

There are some factors, but the most important one would be multisampling. If you know your way around gaming, you probably heard of anti-aliasing. To get rid of jagged edges, you take samples of many pixels around one pixel to determine its accurate color and brightness. Aliasing is not really the same thing for video, but the main mechanic works similar. Where a 1080p recording only has one sample for each pixel in a 1080p sequence, 4k footage on the other hand will have much more information for each pixel when downsampled to 1080p.

  • Surely that can't explain such a huge difference in sharpness, especially since they are at the same bit rate? 4K footage should have 4 times the resolution, but when downsampled, the difference shouldn't still be 4 times? And the thing is, the original 1080p footage looks fine to me, and the downsampling benefit of 4K only appears in lower bitrates. I can't comprehend what happened. – Michael Jun 13 at 14:54
  • I think my main confusion is why the benefit of downsampling 4K is only apparent at low bitrates? I'm concerned that I might be rendering my videos incorrectly. – Michael Jun 13 at 15:01
  • The same amount of information is stored for a pixel no matter what resolution it is: 480p, 720p, 1080p, 4k, etc. Of course bitrates and chroma subsampling makes this much more complicated to understand. I have explained how chroma subsampling is playing a role in this quality difference in my answer. – TravisO Jun 13 at 18:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.