Question Can upscaling a disc ever lookas good as playing it back at its native resolution?

Page 2 - Seeking answers? Join the What HiFi community: the world's leading independent guide to buying and owning hi-fi and home entertainment products.

Arron

Well-known member
Aug 24, 2021
157
51
1,670
Visit site
I think you might need to explain this, because the eye test would appear to contradict it. How is it impossible for content that has been upscaled to look worse than content displayed at its native resolution?

...and just to clarify, in my original discussion, "native resolution" did not mean displayed at native resolution on a 4K screen as a little image in the middle. It meant displayed at native resolution on an equivalent sized screen whose resolution matches the source.
Attached is a zoom on Peter Fonda's jacket from the Easy Rider DVD. One in native and the other an upscale from it to 4K. Exactly the same disc; one without upscaling and one with. It's what happens with bicubic upscaling. There are mathematical reasons why it appears to find detail where none exists — what it's really doing is finding detail that exists but the human eye cannot perceive it.

Bicubic is a fairly standard upscale. Most higher quality players have more sophisticated upscales that will dig out even better images. There are people who spend their entire career on improving upscalers.


Easy Rider.png
 

Corpus_Chain

Well-known member
Jan 15, 2024
39
7
45
Visit site
Attached is a zoom on Peter Fonda's jacket from the Easy Rider DVD. One in native and the other an upscale from it to 4K. Exactly the same disc; one without upscaling and one with. It's what happens with bicubic upscaling. There are mathematical reasons why it appears to find detail where none exists — what it's really doing is finding detail that exists but the human eye cannot perceive it.

Bicubic is a fairly standard upscale. Most higher quality players have more sophisticated upscales that will dig out even better images. There are people who spend their entire career on improving upscalers.


View attachment 6362

The one on the right is the upscaled version?

What TVs are they being displayed on?
 

Corpus_Chain

Well-known member
Jan 15, 2024
39
7
45
Visit site
I don
They're the image that gets sent to the TV from the player.
I think you misunderstand me. Here is a diagram of what I'm talking about, and I was trying to ask whether your test was done according to Scenario A or Scenario B. It's not necessarily important what you did, but for the purposes of understanding this whole thread and my conundrum, this here post will help to explain the scenarios I am discussing.

In all scenarios, it doesn't matter what processing is done by the player or the TV, as the hypotheses are agnostic about this, and the assumption is that the TVs and players are identical in all parameters other than those specified in the diagram.*

In A1, the non-upscaled DVD is not being shown at native resolution, as it is being stretched to fit the larger screen. Obviously this would look terrible.

In B1, the non-upscaled DVD is being shown on a TV that matches its resolution, so no stretching or upscaling is needed.
My hypothesis is that B1 has the best** picture of all four scenarios, as no processing is being done to the picture.


A2 and B2 are exactly the same, and I hypothesise these have the second best picture, doing some processing to the DVD to make it fit the screen in a pleasing way, but no amount of clever processing will get this 100% correct so that you wouldn't notice a difference from B1.

A1 would have the worst picture, as it has been stretched and the additional pixels have been filled by whatever method TVs use for this.

*Obviously, in practice, any SD TV would be much older than comparable 4K TVs, and therefore 4K TVs have modern display technology, so there is a potential separate debate about how this would play out in the real world, but when people upgrade their TVs, they usually upgrade from a TV that is 10-20 years old to a new one, so the thought experiment is valid.

**obviously if you sat too close to the screen, you would be able to see the individual pixels and this would change the result, but at a reasonable distance, this should be better.upscaling problem diagram 1.jpg
 
Last edited:

Arron

Well-known member
Aug 24, 2021
157
51
1,670
Visit site
I don

I think you misunderstand me. Here is a diagram of what I'm talking about, and I was trying to ask whether your test was done according to Scenario A or Scenario B. It's not necessarily important what you did, but for the purposes of understanding this whole thread and my conundrum, this here post will help to explain the scenarios I am discussing.

In all scenarios, it doesn't matter what processing is done by the player or the TV, as the hypotheses are agnostic about this, and the assumption is that the TVs and players are identical in all parameters other than those specified in the diagram.*

In A1, the non-upscaled DVD is not being shown at native resolution, as it is being stretched to fit the larger screen. Obviously this would look terrible.

In B1, the non-upscaled DVD is being shown on a TV that matches its resolution, so no stretching or upscaling is needed.
My hypothesis is that B1 has the best** picture of all four scenarios, as no processing is being done to the picture.


A2 and B2 are exactly the same, and I hypothesise these have the second best picture, doing some processing to the DVD to make it fit the screen in a pleasing way, but no amount of clever processing will get this 100% correct so that you wouldn't notice a difference from B1.

A1 would have the worst picture, as it has been stretched and the additional pixels have been filled by whatever method TVs use for this.

*Obviously, in practice, any SD TV would be much older than comparable 4K TVs, and therefore 4K TVs have modern display technology, so there is a potential separate debate about how this would play out in the real world, but when people upgrade their TVs, they usually upgrade from a TV that is 10-20 years old to a new one, so the thought experiment is valid.

**obviously if you sat too close to the screen, you would be able to see the individual pixels and this would change the result, but at a reasonable distance, this should be better.View attachment 6365
What I did was intercept the bitstream between the player and a 4K TV. Kind-of your scenario A but not exactly.

DVD ----> Player ----> --[intercept]--> TV

Your A1 and B1 will look the same — both will look like my left-hand image capture.
Your A2/B2 is what was in my right-hand image capture — the player did a bicubic 4K upscale of the same image as A1/B1.

I manually did a 2x zoom on the two image captures so you could see the difference more clearly.

The upshot: your A2/B2 will have a much better image — like my right-hand capture. You'll get sharper lines, smoother gradients, etc.

For ref, my main system is:
DVD/BD ----> Plex Server ----> Apple TV ----> Projector ----> 106" screen

The upscaler in the Apple TV is one of the best. It makes DVDs a very pleasant watch even on a screen that big.
 

Corpus_Chain

Well-known member
Jan 15, 2024
39
7
45
Visit site
What I did was intercept the bitstream between the player and a 4K TV. Kind-of your scenario A but not exactly.
Aha, we get there eventually. While your test is interesting as a test of the benefits of upscaling technology, it is not the a test of my hypothesis. It therefore neither proves nor disproves the hypothesis. It is actually very difficult to test the hypothesis, but thanks for posting. Upscaling technology is indeed very clever but it is there to solve the problem of having all these different resolutions being fed into a high resolution TV, which is a problem that simply wouldn't need solving if you had the right TV for the resolution you're trying to display.
 

Arron

Well-known member
Aug 24, 2021
157
51
1,670
Visit site
Aha, we get there eventually. While your test is interesting as a test of the benefits of upscaling technology, it is not the a test of my hypothesis. It therefore neither proves nor disproves the hypothesis. It is actually very difficult to test the hypothesis, but thanks for posting. Upscaling technology is indeed very clever but it is there to solve the problem of having all these different resolutions being fed into a high resolution TV, which is a problem that simply wouldn't need solving if you had the right TV for the resolution you're trying to display.
No, your hypothesis is bunk. Your A1 and B1 will have no visible differences except due to the quality of the panels. Both A1 and B1 will look like my left-hand capture.

There is no visible problem with the a simple stretch of DVD resolution to a 4K screen. You can just about see the effect if you use an SD test card and know what to look for but it's not visible during normal video playback. Even then, a modern upscaler does not use a simple stretch so an SD test card will *always* look better on a 4K TV.

If you truly believe your hypothesis, by all means go out and buy an old 32" SD TV so I can tell you, "told you so" afterwards :p
 

Corpus_Chain

Well-known member
Jan 15, 2024
39
7
45
Visit site
If you truly believe your hypothesis, by all means go out and buy an old 32" SD TV so I can tell you, "told you so" afterwards :p
Hahaha.
Okay, I think I see your point. Well, it's an interesting thought. Now I'm annoyed that my TV doesn't have some way to turn off upscaling!
 

Arron

Well-known member
Aug 24, 2021
157
51
1,670
Visit site
Hahaha.
Okay, I think I see your point. Well, it's an interesting thought. Now I'm annoyed that my TV doesn't have some way to turn off upscaling!
You can see the difference using VLC on a computer. It has a way to turn the upscaler from effectively off to reasonably high quality -- which is what I did.
VLC > Tools > Prefernces > Video > Show Settings All > Video > Filters > SWScale

Bicubic is its best quality -- not as good as the one in the Apple TV but pretty reasonable
Nearest Neighbour -- a simple stretch to fit the screen

Did the screen captures on the 4K screen I use for video editing.
 

TRENDING THREADS