Steve Yedlin, ASC on Resolution

[David Litchfield]

From an 8k source, top is 2k upsampled to 4k, bottom is 4k downsampled from 8k. I think it's pretty obvious which I prefer and which looks better....

But the argument being made is a straw-man. It's also problematic to think of the audience as the only consumer of the image.

Smaller pixels are noisier - unless they are not. To state that as a general rule would be incorrect. It's also incorrect to assume that people in industry don't test to see the benefits and trade-offs of compression / resolution because I know that they do, and they go off the results of those tests and not headline figures.

Graeme

But is this not more noticeable in a still than a moving image?

 

[Joel Arvidsson]

Phils image says it all..next subject!

 

[paulherrin]

But is this not more noticeable in a still than a moving image?

no, the differences are even more pronounced in a moving image

unfortunately, such comparisons would be even more impacted by video delivery compression, so it's just way easier and better all around to deal with a still format (unless you're in an ideal theater setting)

 

[David Battistella]

I want to introduce one more concept here and that is deep focus and compression.

As I listen and read conversations about compression and image quality, I have to say, that if you are shooting shallow depth of feild in low light compression becomes an important factor in noise reduction, but generally, if you are using wide open apertures with blurred backgrounds, not a lot of camera movement or image detail, in a normal lighting situation, why do people feel the need to capture at 5:1 compression?

Because it is better, or because they think it is better?

On things like interview footage, with a camera on a tripod and large apertures, I have learned that you can easily compress the images way beyond what you would expect without any real loss in detail or skin, or post flexibility.

David

 

[Misha Engel]

Misha,

You are missing some fundamental information.

It is called REDCODE RAW. They are not two separate things REDCODE IS RAW when you are talking about RED. The whole reason you can record RAW to a MINIMAG is because of REDCODE RAW.

REDCODE RAW is a variable compression RAW format. If you record it 10:1 or 2:1 it is still RAW.



David

Redcode Raw works as stated by Graeme

Compression is done in three stages - transform, quantization and entropy coding. If you're doing mathematically lossless you omit the quantization step. You can also take any lossy compression and turn the quantization off and see what compression ratio you get. Mathematically lossless compression is always variable rate as it is entirely content dependent. It's only by going to lossy compression can you target a fixed data rate.

So what happens in RedCode - you should see increasing file size per frame as you lower the compression. The point at which the file doesn't get any bigger is the point where you've reached mathematical losslessness. That will occur somewhere between 2:1 and 3:1 and is image content dependent.

Graeme

Redcode Raw is processed compressed RAW, at 2:1(3:1) it is mathematical lossless, at a higher compression rate it is lossy.
The Helium records 8k 24fps at 5:1 and upwards, so it is lossy (very good lossy, but lossy).
At 8k 60 fps the compression goes up to 12:1, even more lossy (still good lossy).

So I get the fundamentals.

All I'm asking for is that DSMC3 (or whatever the name will be in the future) will have faster media so that the oppertunity is there, to produce a mathematical lossless picture.

 

[Nathan DuMoulin]

All I'm asking for is that DSMC3 (or whatever the name will be in the future) will have faster media so that the oppertunity is there, to produce a mathematical lossless picture.

He wants faster data rates, to achieve lossless 2:1 compression at 8K. Makes sense, and is a reasonable request.

Now that the whole 1:1 confusion is out of the way, let's get off this tangent.

From an 8k source, top is 2k upsampled to 4k, bottom is 4k downsampled from 8k. I think it's pretty obvious which I prefer and which looks better....

The intended example mentions one being upsampled while the other downsampled, however at 360p, they both end up downsampled. Did I miss the full size version somewhere, or am I missing something else?

 

[Michael Tiemann]

I want to introduce one more concept here and that is deep focus and compression.

As I listen and read conversations about compression and image quality, I have to say, that if you are shooting shallow depth of feild in low light compression becomes an important factor in noise reduction, but generally, if you are using wide open apertures with blurred backgrounds, not a lot of camera movement or image detail, in a normal lighting situation, why do people feel the need to capture at 5:1 compression?

Because it is better, or because they think it is better?

On things like interview footage, with a camera on a tripod and large apertures, I have learned that you can easily compress the images way beyond what you would expect without any real loss in detail or skin, or post flexibility.

David

Would be cool to add a compression loss bar to complement over- and under-exposure goalposts. REDCODE engine could report how much leftover bandwidth it thinks it has at a given compression level. For busy scenes it might port lossiness at 5:1. For simple scenes with lots of soft focus, it might report lots of headroom at 12:1.

 

[Graeme Nattress]

Another aspect we've not talked about yet is HDR. I've been working on some HDR stuff this last week, so while I was testing that, I also took a look at resolution with respect to dynamic range. It's well known that contrast plays a big part in the perception of resolution in that low-contrast details are harder to see than high contrast details and that sharpening works by increasing edge contrast.

HDR allows for display of higher contrast imagery and thus can make high resolution easier to see. Conversely it can make it easier to determine soft images as soft, so resolution will play a bigger role as we move towards higher dynamic range displays.

Graeme

 

[Michael Dalton]

i was suprised by Alexa65 which looks sharper and better than imax 15perf 65mm.

You would have been more amazed to see a camera comparison test we did for Giant Screen and IMAX in Sept comparing many cameras with Helium and Vista. It was clear that Helium and Vista were the winner above all. F65 looked good, but inferior to 70MM 15 perf. BM and Arri Alexia we're unusable. Vista had a quality that beat helium, but marginally.

I did several test where I digitally extended shots to get a True 8K (truer as it was 8192 x 6144 and not 8192x8192 which is true 8K) which looked way better vs crop ins that are essentially true 4k with a vertical resolution near 4k. Until we get a full frame 4:3 sensor IMAX that shoots something like 8192 x 6144 to do IMAX on a Red will require a lot of painstaking and costly work to get there.

For anyone who says resolution is irrelevant, my argument is they must be content with HD. I think Red along with most of the users here are interested moving technology forward. We can see the difference when viewing different resolutions (8k,6k etc) so to view this test at 2k is a bit pointless. It's funny that this is a non argument in the stills world, you don't hear Photographers saying 12mp is perfectly fine. They want better smaller pixels.

For sports or nature, IMAX and or epic films, the added detail can only enhance our story telling. Maybe because we can get away from the 2 second shots cuts of film and let stuff breath a bit.

It does not matter to the general audience?? haha, put an HD TV next to a 4k and 8K Tv and ask the general audience what they prefer? I'm sure there would be a clear winner.

With Cinema stagnating in the 2k world, are they not jeopardizing their future against NETFLIX and HBO where they can tell bigger stories that that look spectacular in 4k but can fill the human crave for more?

 

[Michael Dalton]

One thing I'm curious is if Red is looking at all these 70mm film and planning to make a camera for them? As mentioned in my previous message that I feel film is stagnating, JIM, Jarred and Graeme, if you want to turn the film world upside again, maybe the future is BIGGER. Lets give the audience something incredible. We're getting a taste of how epic 70mm is again, maybe we can shift hollywood there. If you guys have not been to IMAX movie in a while, go check one out (just make sure it's shot in true IMAX and not some passable 1;85;1 format. If you have not been to DOME Omnimax movie ever, then I suggest you trying that! It's an incredible format. Graeme, we have one here in Toronto at the science centre! If your in town and want to take in a film send me a message (keep in mind that it has to be shot for a dome). If anyone has the chance to see an 8K Dome film (there are only a few in the world) with actual 8K content, it's something truly special.

 

[Michel Hafner]

You would have been more amazed to see a camera comparison test we did for Giant Screen and IMAX in Sept comparing many cameras with Helium and Vista. It was clear that Helium and Vista were the winner above all. F65 looked good, but inferior to 70MM 15 perf. BM and Arri Alexia we're unusable.

Arri Alexa or Alexa65?

 

[Michael Dalton]

65! the Alexa did not have the resolution to hold up, so they have been out of the equation with BM a long time ago. IMAX has struggled because it's a resolution monster, and because of budgets people have looked to alternatives to 15 perf film (which used to be accepted, but has been killing the industry). Even today, unless your willing to put the effort into the extensions digital does not compete with 15 perf 70mm. Remember IMAX is 3D medium, not 2D. It's a lot of data and work.

I must say 3D 8K dome, is mind blowing!!! E&S did a demo of some CG stuff and it was incredible. I can't wait to see my first film test for my next film shooting in native 8K 3d for the dome.

 

[Patrick Tresch]

The problem is OVERSHARPENING!

One key point in Mr. Yeldin observation is :

"The perception of clarity and sharpness is actuall increasing when the true resolution is decreasing" this is the argument to tell the talent : more resolution is smoother images.

IMHO Alexa65 (and XT and F55) is over sharpened (or the scaling algorithm he is using) and on the Weapon6k material (due to strong OLPF?) is to soft.

Now I've never seen such a strange sharp green noise structure in a R3D. Would love to see the R3D to play with it.





You don't get the same results uprezzing a 2k picture to 4k when it's first generated from a 6k RAW file than from a 3k RAW file. But at the end... in the same technical pixel peeping logic : who cares when the narrative is good?

 

[Mike Finegan]

All sounds speculative - Still frames are not the same 'resolution' to the human visual system as a series of moving images. The mind interpolates inn time (along with every other dimension).
So the OLPF of the DSMC2 is noticeable in compressed still frames (even 2:1 compression) but can be perfect - lossless and sufficiently detailed as moving images even at relatively high compression rates.
I once heard an IMAX post manager say the IMAX preference is for color depth then frame rate and only finally resolution. He claimed they inerpolate up routinely.
I would like to see film makers use lowest possible compression and high sharpness lenses for cinematic landscapes that are supposed to be in focus

 

[Tom Dowler]

I agree with Yedlin, but this is the definition of an appeal to authority fallacy. Yedlin's position is well defended not because of the work he's done making movies, that is actually completely irrelevant. His position is well defended because he presents evidence of his claims.

busted.


But yes, his position is well defended. And his work and reputation are stellar. I'm not saying he's right because of who he is or what he has achieved. He's right because his argument is well defended. His detractors here have poorly defended arguments, but are also not people noted for their expertise.

I don't think it's an appeal to authority to prefer an expert's opinion over someone with less expertise. Going down that road gets you into dangerous "I don't trust experts/scientists" Trump/Brexit territory.

 

[Misha Engel]

65! the Alexa did not have the resolution to hold up, so they have been out of the equation with BM a long time ago. IMAX has struggled because it's a resolution monster, and because of budgets people have looked to alternatives to 15 perf film (which used to be accepted, but has been killing the industry). Even today, unless your willing to put the effort into the extensions digital does not compete with 15 perf 70mm. Remember IMAX is 3D medium, not 2D. It's a lot of data and work.

I must say 3D 8K dome, is mind blowing!!! E&S did a demo of some CG stuff and it was incredible. I can't wait to see my first film test for my next film shooting in native 8K 3d for the dome.

Sony F65 or Arri Alexa65 ?

 

[David Battistella]

Would be cool to add a compression loss bar to complement over- and under-exposure goalposts. REDCODE engine could report how much leftover bandwidth it thinks it has at a given compression level. For busy scenes it might port lossiness at 5:1. For simple scenes with lots of soft focus, it might report lots of headroom at 12:1.

This is a cool and useful idea. Unfortunately there will be people who don't understand compression that will insist on 1:1. I for one do not see a RED with a CODEX box attached in the future, but that is just me.

David

 

[Julian Banos]

I absolutely agree. I did a test with people who are not involved in the film business, I asked them to step inside a color grading suite, on the right side we had a Panasonic Plasma display calibrated to 709. On the front we had an LG OLED calibrated to DCI P3. We were playing from two different apple TVs CHEF S TABLE FRANCE one was playing regular HD NETFLIX and the other one DOLBY VISION. The 15 person focus group watched for one minute the HD program, then for one minute the 4K HDR program and back to 2 minutes on HD and back to 2 minutes on 4KHDR.
Everyone expressed three adjectives superior to HDR 4K: Color, Definition, Depth. 12 out 15 expressed a greater sense of light range (dynamic range).

We have done the same with regular 4K vs HD and most people saw the difference, but not all. The response for resolution on HDR is there. It is really simple to conduct the same test.

 

[Björn Benckert]

The problem is OVERSHARPENING!

One key point in Mr. Yeldin observation is :

"The perception of clarity and sharpness is actuall increasing when the true resolution is decreasing" this is the argument to tell the talent : more resolution is smoother images.

IMHO Alexa65 (and XT and F55) is over sharpened (or the scaling algorithm he is using) and on the Weapon6k material (due to strong OLPF?) is to soft.

Now I've never seen such a strange sharp green noise structure in a R3D. Would love to see the R3D to play with it.






You don't get the same results uprezzing a 2k picture to 4k when it's first generated from a 6k RAW file than from a 3k RAW file. But at the end... in the same technical pixel peeping logic : who cares when the narrative is good?

Also difference in focus, Not really on the eye on that image, more on the nose. And I assume exposure and lens setting is the same for all cameras, Dragon underexposed as this image clearly is create noise that also gets amplified by the compression and digital gain.
Dragon with better focus and exposure to not be so close to the noise floor would render a total different result.

Here is a epic MX eye... So not dragon as i the test above, only 5k MX, but with exposure set to avoid noise, and focus is somewhat on target. Not the best lens but decent. So most other parameters set so that resolved resolution is not drowning to deep into other factors.

A006_C024_12197X by Björn Benckert, on Flickr


Crop 1:1 from original 5k image: 1to1ropMX by Björn Benckert, on Flickr

 

[PatrickFaith]

Another aspect we've not talked about yet is HDR. I've been working on some HDR stuff this last week, so while I was testing that, I also took a look at resolution with respect to dynamic range. It's well known that contrast plays a big part in the perception of resolution in that low-contrast details are harder to see than high contrast details and that sharpening works by increasing edge contrast.

HDR allows for display of higher contrast imagery and thus can make high resolution easier to see. Conversely it can make it easier to determine soft images as soft, so resolution will play a bigger role as we move towards higher dynamic range displays.

Graeme

I saw a real good example of this recently on a immersive setup where there was a book with text in a shadow with a light shown on it. I just don't think people think about it, but text shows up like crap in films & tv, so people are currently shooting assuming that text should not be readable in most cases(aside from big signs and stuff). But in the example I saw recently, with the increased pixels I was still having a hard time reading it ... but when the 'hdr' algorithms kicked in the text popped and I didn't have any eye strain.