Simple way to see if you have 4K ( no UHD TV needed )

[Les Dittert]

Here is the easiest way to tell if your footage is really 4K :

1) Look at it sized down to HD on an HD monitor. ( like 95% of user deliver to clients right now )
2) Push in on it 200% ( making it twice as big )
3) Look at the pushed in version on the HD monitor.
4) Does it still look nice and sharp ? If so, you have some real 4K there.

 

[Stephen Williams]

Here is the easiest way to tell if your footage is really 4K :

1) Look at it sized down to HD on an HD monitor. ( like 95% of user deliver to clients right now )
2) Push in on it 200% ( making it twice as big )
3) Look at the pushed in version on the HD monitor.
4) Does it still look nice and sharp ? If so, you have some real 4K there.

Well when I punch in 1:1 on my Epic, the image looks soft & noisy, so I don't think your theory is very useful.

 

[Les Dittert]

Well when I punch in 1:1 on my Epic, the image looks soft & noisy, so I don't think your theory is very useful.

Perhaps you don't have 4K detail in your '4K' images.

The basic way some people describe UHD monitors is that it is like 4 tiles of HD ( which is correct ).
Each tile is capable of displaying a nice HD image on it's own with no compromise or fuzziness.
Pushing in 200% is the same thing, technically.

 

[Stephen Williams]

Perhaps you don't have 4K detail in your '4K' images.

The basic way some people describe UHD monitors is that it is like 4 tiles of HD ( which is correct ).
Each tile is capable of displaying a nice HD image on it's own with no compromise or fuzziness.
Pushing in 200% is the same thing, technically.

Shoot a bank note, that has plenty of detail, I suspect what your looking at does not have 4k detail so you don't see it.

You just need to shoot 4k & 2k cut together viewed on an HD television & you will see that 2k is quite soft relative to 4k.

 

[Les Dittert]

Ummm, you are the one with 4K detail issues, not me
I shoot sharp 4K and go through great lengths of effort and image processing to make sure.
( no OLPF and special processing to remove aliasing and moire artifacts to make up to 6K images with no Dragon sensor )

Do you agree that 4K is like shooting 4 HD tiles at the same time, so that each HD ( call it 2k ) should be just fine looking in an HD monitor ?
Pushing in 200% is the same exact thing, or for that matter cropping out a 2K 'tile' should look sharp as well.
If you have 4K worth of sharp image.

 

[Stephen Williams]

Ummm, you are the one with 4K detail issues, not me
I shoot sharp 4K and go through great lengths of effort and image processing to make sure.
( no OLPF and special processing to remove aliasing and moire artifacts to make up to 6K images with no Dragon sensor )

Do you agree that 4K is like shooting 4 HD tiles at the same time, so that each HD ( call it 2k ) should be just fine looking in an HD monitor ?
Pushing in 200% is the same exact thing, or for that matter cropping out a 2K 'tile' should look sharp as well.


If you have 4K worth of sharp image.

4k bayer is not 4k as we all know very well, so I don't expect a 2k Bayer image to resolve 2k. Alexa uses down samples to make HD. So I do not agree that a 4 k image cropped to 2k will be as sharp as an image intended for 2k presentation or from a 3 CCD HD chip camera.

 

[Phil Holland]

Remember that Red cameras record a raw data stream into a user compression defined codec called REDCODE RAW that gets debayered on the back end.

Two things on that. A mathematical "perfect" 4K debayer is creating by shooting at a higher resolution, such as 5K and scaling the image down. Typically you multiply your captured image resolution by 80% to find your true debayer resolution. However, this doesn't take into lens quality or the OLPF for instance. And more importantly it doesn't take account for any sort of user guided post processing.

Raw data may not appear "shaving sharp" as captured. However, you have control over that look in post via various sharpening tools and techniques. There is a great deal of logic behind this as you may not always want needle sharp images and may want a more organic "film-like" feel. Which, if you do side by side comparisons of 4K and 5K Red camera material why it has a similar feel to film. REDCODE RAW's wavelet compression scheme is a bit part of that look. Additionally, because it's wavelet based and the files are rich you can extract a great deal of detail from your captured material. Also, it makes it much easier to mix and match resolutions via scaling, such as upscaling 3K to a 4K timeline.

Now removing the OLPF can create sharper out of the camera images, but as Les has mentioned some additional processing is required to remove aliasing and moire artifacts. I don't totally see how he's resolving 6K from a true 4K image, but I know of several methods to capture more detail via super sampling and "creative" upscale and sharpening methods. There's a Epic with the OLPF removed out in Woodland hills and we did some tests back in 2011 to see what that yielded. However, it was more or less a wash as with modern sharpening algorithms we're getting a great deal of detail out of the REDCODE RAW and the overhead added by post processing the artifacts out was less desirable. However, if I was hell bent and determined, perhaps shooting a landscape piece, I could see some fun advantages to removing the OLPF and some of the "crazy stuff".

The real question throughout all of this is what exactly do you want from your final image. Also, perhaps you may want your close ups to be more on that organic side and maybe your wides on the "ouch that's sharp" side. I tend to balance the two mindsets out and not have extremely ruthlessly sharp images, but enough to reveal detail and subtle texture. For narrative I like an edge typically, but one that feels like it's been through a few sword fights first.

 

[Phil Holland]

I should also add, bringing the Epic Mysterium-X Monochrome into this discussion also provides a higher resolved image.

 

[Les Dittert]

Phil,
Back to my original point: How do you like my 'acid test' for 4K results , that is it should look like full HD if you crop into it and view on an HD panel.
UHD's mantra is that it is equivelent to 4 simultaneous HD feeds in visual effect. Four 1920x1080 tiles = one UHD 3840x2160.
If I crop out one of those four HD tiles from UHD footage, it should still look like a 'regular' HD program with no excuses.

Kinda like a checkmate situation for people claiming to shoot 4K.

 

[Matt W.]

Even if your sensor/OLPF combination has consistent 100% MTF to 4k (yeah right) you'll still get more noise and your lens's MTF will surely be falling off by then. No camera is giving you this kind of image quality at this point. 4k remains a buzzword... Arri never promoted 4k film scans as claimed for their improved sharpness but rather for reduced aliasing when capturing the "about 2k" image most film/lens combinations resolved to a decent MTF.... Remember we're at 30% MTF at best if you believe the "3.2K" figure for 50D film... Area under MTF curve for the same film is less than that under the Alexa's MTF curve I'd bet, too. Epic probably sharper still, but I remember the original red was much softer than most 1080p cameras from the same time... way too aggressive on the OLPF until the MX.

Yes 4k delivery is 4x 1080p... Content is not even close. This is one of many reasons why 4k will remain a buzzword for a long time, at least in home video. Might catch on theatrically pretty fast.

 

[Phil Holland]

Phil,
Back to my original point: How do you like my 'acid test' for 4K results , that is it should look like full HD if you crop into it and view on an HD panel.
UHD's mantra is that it is equivelent to 4 simultaneous HD feeds in visual effect. Four 1920x1080 tiles = one UHD 3840x2160.
If I crop out one of those four HD tiles from UHD footage, it should still look like a 'regular' HD program with no excuses.

Kinda like a checkmate situation for people claiming to shoot 4K.

Well it's a nice way to go about it and an easy way to "see" what you're supposed to be seeing in terms of detail. That's nothing new though, that's what we've been comparing to since the Red One MX's early days.


There are other variables always at play when it comes to shooting material and finishing material. There's the technical approach and the artistic approach. I can shoot and finish in 4K, but I can use lowcon, nets, and heck even Vaseline on the lens and still state that's it a 4K production. That certainly won't resolve 4K on a good day. However, the aesthetic may fit what I wanted out of the shot, scene, or production.

The technical approach comes down to numbers. To get a real number you need to shoot a standard resolution chart and also take into account the lenses used in that test as well (and that should be a longer lens for that type of test typically in the 50-100mm range). Particularly I was curious why the same lenses weren't used in your film against Scarlet-E test. Without a control there's no true 1:1 comparision there. Real world tests like that are great though and valid. However, to "know" and have a hard technical answer it's gotta be consistent and something people can recreate.

The last resolution test I did we shot on a single Leica Summilux-C prime. That control led to reveal specific results and across the 6 cameras tested we used the same glass, same exact working distances, same lighting, same charts, etc....

That's why we know what the Epic and Scarlet resolve in terms of line pairs with and without the OLPF and that's also how we measure lenses. Consistency and standards.


But, yes, generally speaking a 1/4 frame of UHD 4K (3840x2160) material can look as good as a full frame high quality 1080p (1920x1080) image. However, if the whole image will look that way? That depends on subject matter, focus, aperture, etc.... If a portion of your frame is out of focus, it's just out of focus. The allure of 4K comes from the detail that's captured within the frame or a portion of the frame. To some it provides more depth and richness.

 

[Les Dittert]

Here is my content. This 6K frame was shot at 4K on a Scarlet.
http://www.mediafire.com/download/8hfytb52qjjgofb/2-10_P5fr6K.jpg

I can easily get four 1920x1080 out of this.

Shot on a non-L canon 17-55 lens.

 

[Phil Holland]

Here is my content. This 6K frame was shot at 4K on a Scarlet.
http://www.mediafire.com/download/8hfytb52qjjgofb/2-10_P5fr6K.jpg

I can easily get four 1920x1080 out of this.

Shot on a non-L canon 17-55 lens.

This is primarily asking why you feel this is a 6K image:



*click to open in new window for the 100% crops.

I see edge artifacts/aliasing, destructive scaling or sharpening (there are halos around contrasty areas), and there's moire all over the image.

I can also see this when scaled down to 4K on this particular image.

Additionally when applying standard chroma blurring to the image you can remove a bit of the artifacts, but those edges are still funky pants:

Even with this technique it didn't solve all of it.

 

[Les Dittert]

Maybe asking it to be all the way up to 6K is a bit of a stretch

How about this retuned version, at 4K and shot at 4K.

Here is a 1:1 crop near the center:

and then the area which looks worse, closer to the edge:

I posted 1:1 because that is what you would see on a fine pitched monitor.
There is some roping on a few of the edges, but honestly when it is reduced further still for a 3840 UHD TV, it looks great in motion.
When you are close to what I call the 'pixel on - pixel off' type of resolution, you are always going to be seeing some artifacts because you are in the danger zone of sampling theory.
But I like the zap of the image, so to speak.

My point is still : UHD is quad HD, so if you crop into UHD to get 1920x1080 media, it really should look no worse than you would expect an HD show to look like.
Not often the case, but it can be done.
Not all media is supposed to be razor sharp, but some of it better well be, or what's the point of the 4K mantra.

 

[Phil Holland]

Looks a bit better. I can see a bit of moire on some of the "frames", but there's ways to deal with that. On the extreme chroma blurring some color resolution/detail is lost.

Just curious, are you doing these tests at REDCODE 3:1?

 

[Björn Benckert]

Perhaps you don't have 4K detail in your '4K' images.

The basic way some people describe UHD monitors is that it is like 4 tiles of HD ( which is correct ).
Each tile is capable of displaying a nice HD image on it's own with no compromise or fuzziness.
Pushing in 200% is the same thing, technically.

But Les, that would mean that you have no noise what so ever in your shots... If so, please upload such frame I would be curious to see. As I never seen a Cmos image with no visable noise and as far as I know downsampling makes less noise. So when you downscale a "perfectly sharp" 4k image the texture of the image will improve as the noise gets down sampled. when you look at the same image at 1:1 than that effect does not take effect.

So basically every 4k image I have seen looks as of less quality even seen with your eye at 1:1 compared to 2:1 downsampling. That is even true for a non debayer camera as your pixel raster then gets super sampled.

So I would say, your theory makes somewhat sense but it's pretty far from the whole truth.


You also have to consider, there are no perfect lens. There are sharper lenses and there are lenses that are not so sharp. But there is no lens that is perfectly sharp... So that also means that down sampling is helping bringing back sharpness lost trough the glass. As a downsized picture is always holding more info per pixel than it's original bigger version, yes?

Edit: saw your frame now, and I saw it before. It's nice in some ways. But some might say it looks like a High F-stop, Post sharpen, quite low quality image.

To me lens resolution is one thing. I can actually debate that I do not really like super sharp lenses. (It's actually one reason why I spent some money on Hawk's) But I do however like pixel resolution, that is something completely different, you basically cant have enough of it even if you shoot complete blured out lights or such, the extra pixels will show... Or not show might be a better term.

I see pixels clearly on a normal 2k DCP in a normal sized cinema no matter if it's pixel perfect graphic elements showing or blurry faces . At 4k, I don't. Thats the important difference....

Captured measured 4k resolution might be nice for some things, but it does not mean that your image is more 4k than a blurred image.


Also where are you zooming in and out between 1:1 and 2:1 ? if it's done in the viewer in red cine x pro you only see a very rough / "nearest" kind downsampling. The downsampled image would be it's best first when rendered out as rescaled. Your 6k frame above would probably benefit heavily if downscaled to 3k with lancros or sinc downsampling. Then some of the quite heavy artifacts you got along diagonal edges and such could probably be rescued.

 

[Les Dittert]

Phil, it was 8:1 Scarlet. I wish it was 3:1.

I think if I added some grain it would look more 'organic' ( I'm not a fan of the word ) and therefore more pleasing to many.

Looks a bit better. I can see a bit of moire on some of the "frames", but there's ways to deal with that. On the extreme chroma blurring some color resolution/detail is lost.

Just curious, are you doing these tests at REDCODE 3:1?

 

[Phil Holland]

Phil, it was 8:1 Scarlet. I wish it was 3:1.

I think if I added some grain it would look more 'organic' ( I'm not a fan of the word ) and therefore more pleasing to many.

Moving forward with your tests on Scarlet-E, I would suggest shooting at 6:1. You may see a visible bump in what you're doing especially since you took out your OLPF. I pretty much shoot 5:1 on Epic 90% of the time and 3:1 for high frequency detail scenes. I tend to dip into other compression ratios mostly for over cranking.

Here's a link to a UHD 4K "finished" image with what I refer to as "Print Sharp Fine Art Paper" derived from a 5K HD frame, I think 75mm T4 5:1.
http://www.artbyphil.com/temp/reduser/forLes_UHD4K_PaperGrain5K4KFin.jpg
(*note the .jpg compression may effect the appearance of the grain, but the general idea is there)

This particular look is sharp and has a fine grain pattern that looks rather nice when projected. It's a subtle fine grain and not exactly too film-like, but more of a light textured paper pattern. I would use something like this for nature or fine art work.

Over the years working with .R3D files I created a bunch of different sharpening presets in various applications as well as grain patterns. I've also gone a bit further into an upscale and sharpen to down sampled image technique which provides "crispy" results without jagged edges, but I would say is ruthlessly sharp with the right glass. Depending on what effect I want on the image, that's if I want something that's out of the norm at all really, I'll dig in or create a unique preset for a project, shot, and sometimes even on a per lens basis if I'm dealing with softer glass or attempting to visually match optics of varying rendering qualities.

For most folks who are trying to achieve a film-like look I mostly recommend a simple down sample from 5K to 4K or straight 4K if they shoot that. REDCODE looks rather organic on it's own to me, but if you carve into it deeply you can see the nanobots that create it. So I try to shave and not cut too deep mostly.

Next time I'm out and about and have a nice detailed scene I'll shoot a test and perhaps show some variations on finishing.

 

[Phil Holland]

I should also add that results like what I just posted above would be damn near impossible on traditional 4-Perf Full Aperture scanned at 6K and sampled down to 4K on a Northlight scanner.

Which is why some feel already that if shot carefully 5K on the Epic with Mysterium-X sensor is penetrating into the 65mm/VistaVision territory in terms of resolving power.

 

[Karim D. Ghantous]

as far as I know downsampling makes less noise.

I don't agree, based on my experience. Downsampling preserves noise (or, more accurately, noise characteristics). Downsampling does produce an image which is sharper than native (duh!), while at the same time having smoother noise. Not less, just smoother.