Is 5K the look around for 4K___does 4K look better than 5K scaled down?

[Martin Stevens]

I was just reading a very intelligent and revealing web article by Michael Cioni.

see...

http://michaelcioni.tumblr.com/

Do others here on RedUser agree that a 4K extraction looks slightly better than
a 5K to 4K down-conversion?

I have been thinking of shooting 5K for 4K for a while now. I have Zeiss Standard Speed lenses
and so anticipate that when I get my beloved EPIC-X I will have to shoot 4K due to these lenses
not covering the 5K area properly.

I would love to see some real world examples of Red Epic footage
or stills that reveal this slight aesthetic difference between
a 4K exaction and a 5K to 4K rescale/down conversion.

 

[Mark Andersen]

What I think you asking is do 4K images from a 4K captures look better than scaled down images or 5K captures from the Epic? My instinctive answer would be absolutely NOT. I is generally always better to over sample and scale down. That said there are differences in the quality due to scaling ratios etc. like Quad HD 4K is supposed to scale to HD better than 4096 4K, because it is easier math, it certainly will process faster in RedCine. Then again I don't know everything about digital science, long ago a mentor told me "you don't know what you don't know" I have never forgotten that. Oversampling helps with sharpness and moire problems, the OLPF will soften the image so oversampling hides it. In theory 6K would be even better for a 4K final. I love to hear what others have to say. I'm sure much of this has been covered many times on this forum.

 

[Martin Stevens]

What I think you asking is do 4K images from a 4K captures look better than scaled down images or 5K captures from the Epic? My instinctive answer would be absolutely NOT. I is generally always better to over sample and scale down. That said there are differences in the quality due to scaling ratios etc. like Quad HD 4K is supposed to scale to HD better than 4096 4K, because it is easier math, it certainly will process faster in RedCine. Then again I don't know everything about digital science, long ago a mentor told me "you don't know what you don't know" I have never forgotten that. Oversampling helps with sharpness and moire problems, the OLPF will soften the image so oversampling hides it. In theory 6K would be even better for a 4K final. I love to hear what others have to say. I'm sure much of this has been covered many times on this forum.

I think your view is indeed the convention wisdom regarding this scaling vs non-scaling issue, and has also been
my view for years, but now this web article by Michael Cioni seems to point to a better way.

This is why I wonder if any RedUsers have already done similar tests.

???

 

[Karim D. Ghantous]

That's a 33% loss in resolution. Think about it: a 33% loss. I'm not saying it's wrong - composing loosely is a very pragmatic thing to do. I am not sure that I would do it to that degree, though. However, I'm not a DP.

 

[Martin Stevens]

That's a 33% loss in resolution. Think about it: a 33% loss. I'm not saying it's wrong - composing loosely is a very pragmatic thing to do. I am not sure that I would do it to that degree, though. However, I'm not a DP.

I think the point is that Cioni has empirically seen the extraction look better than the down-conversion.

 

[Dominic Jones]

You need to be very precise when talking about workflows. What Michael is talking about is an extraction from 10-bit RGB (DPX) 5k file vs a scaling from same - not a scaling from the original Bayer (RAW) file vs extraction from same.

I would want to do some serious testing before coming to any concrete solutions regarding an end-to-end workflow. Michael's testing centred around a need to work with RGB DPXs that had been decided upon earlier in the post chain, and therefore may not be representative enough to make a sweeping statement like "it's better to crop than to scale". Certainly in many situations it is better to scale.

As ever, the answer is to test heavily for your preferred solution to every new workflow and deliverable combination you need to encompass - that is undoubtedly what Michael did to arrive at the decision he did for his specific set of circumstances...

Happy holidays,
Dom.

 

[Martin Stevens]

I still wonder if some RedUsers have some tests they could post imagery from regarding this issue?

 

[Erik Franzén]

All things being the same, conventional wisdom tells us that the high-res file scaled to a lower-res format will always trump a recording in said lower-res format. To date, I've not seen a single credible whitepaper, article or technician say otherwise -- and I'd be inclined to disagree with anyone who said so. But as Dominic says, he (M Most) is not talking about scaling from the original RAW file. So if you want to call it fair you have to measure apples to apples.

 

[Frank Cueto]

I have done some testing regarding this, and it DOES provide for a more "Transparent" image; for lack of a better term. It is a texture thing, indeed subtle, but when u do A/B you can appreciate it. Mr. Cioni was one fo the first I remember to advocate the $KHD workflow back in the Plaster City days.

I know we always recommend 4KHD, but usually get trumped by the DPs desire for more FOV.

 

[Patrick Tresch]

I know we always recommend 4KHD, but usually get trumped by the DPs desire for more FOV.

- 4kHD is better for HD 1080 but not necessarely for other formats.
- DPX is perhaps the most common format but limits what you can get from a R3D. There is perhaps no other solution as of today of using DPX in VFX heavy post-production workflow?

On top of this, by doing a center extraction, there turned out to be (what I consider) an accidental benefit: By shooting 5K with a 4K center extraction (or “5K for 4K”), images shot on the EPIC undergo a subtle texture change. The texture becomes a bit smoother because the bayer-pattern pixels are not scaled down from 5K to 4K, rather they are instead cropped to 4K. While still clearly boasting a 4K feel, the subtle difference between a scale and a crop presented what I consider an aesthetic benefit that came somewhat unexpected.

Below is the framing chart for the EPIC on GDT that was created by AE Tyler Nelson. Because EPIC’s native resolution is so significant, one can create their own custom extraction that suits the project. There are numerous technical benefits, such as enhanced look-around, lens millimeters re-line up, older or wider lenses do not vignette, etc. But the slight difference in image texture due to the crop is the one that I think many people will like. There are many ways to get the image texture of any camera to be different based on optics, but when that option isn’t available or ideal, consider this technique which is responsible for the pixel texture you will see in 4K and 2K projections of the film.

Indeed, it would be great to understand this "slight difference in image texture".

I just finished a 4k short film shot in 5k on Epic. We monitored everything in 1920 (1/4) res debayer for the color correction. Making it possible to finish an all 4k project in a indie workflow (Resolve). Wonder if we will run into troubles... I'll teel you later on the 4k Barco. ;-)


Pat

 

[Frank Cueto]

Well, It might not have anything to do with 1080P in Mr. Cioni's case. I imagine that projecting a 4K NON interpolated image must have seen "Less Processed" for a lack of better terms... I can DEFINITELY agree with this logic.

 

[Björn Benckert]

Well, did not read the whole blog but I disagree heavily on :" by doing a center extraction, there turned out to be (what I consider) an accidental benefit: By shooting 5K with a 4K center extraction (or “5K for 4K”), images shot on the EPIC undergo a subtle texture change. The texture becomes a bit smoother because the bayer-pattern pixels are not scaled down from 5K to 4K, rather they are instead cropped to 4K. While still clearly boasting a 4K feel, the subtle difference between a scale and a crop presented what I consider an aesthetic benefit that came somewhat unexpected."

If the epics sensor was twice as big, If it was shooting 20k and being about 6cm wide then I would for sure still use the full sensor and then interpolate down to whatever format to work in. A CMOS sensor does not look to good in 1:1 it kind of needs a 30% downsize with good interpolation. And to me a big sensor/backplane/projection area is always to strive for. Why shoot 70mm neg and crop out a 35mm film in the middle??? Tell anyone that shoots 70mm film and I think they would come after you with a knife. Down sample is good, using many pixels are also good, having a big backplane in your camera is a luxury, so to hear that someday thinks its a great workflow to trough it away by default is crazy. Its off course sometimes better to be on a little bit wider lens, so in post there is some extra to zoom into for stabilization repacking etc. But please try to use the much out of the sensor as possible.

DPX is one way to go, another possible solution is to stay in R3D as long as possible then you can move faster and keep the original files longer with open iso etc. Most workstations takes the r3d today, Do not know about todays Quantel boxes. Personally I hate them, They had their high in the 80's and then I actually sat behind two Editboxes and a HAL for almost 4 years, moved on to fire and inferno made some money and was stupid enough to invest in a Domino... Since then Im ready to puke just from hearing the name..

My tip / five cents on the subject. Get anything else than the q boxes, today premier and resolve outperform them and they cost nothing in comparison, for effect work get a fully fledge flame and/or nuke.

 

[Patrick Tresch]

In Cioni's post smoother means better in an aesthetic point of view. Like using a softer optic. He doesn't mean better resolution.
It's a valid point of view also when almost the whole film was shot in 4k not 5k.

Pat

 

[Sanjin Jukic]

Michael Cioni is an expert and probably he knows exactly what he is taking about.

GDT EPIC CENTER-EXTRACTION FRAMING CHART

 

[Karim D. Ghantous]

Further ramblings based on recent replies

Further ramblings based on recent replies

I just thought of something that I missed before. I said earlier that a 33% loss of resolution - yes, it's 33% - seemed too much. However, the 5K frame is 2:1, not 2.4:1. To crop the 5K frame to get 2.4:1 you have to lose about 427 (2560-2133) lines, or a little over 16% of your resolution. So either way you have to lose something if you want 2.4:1 (I like it but I don't feel an imperative to use it if I were shooting a film).

I would be surprised if RED did not release an 8K camera next year. I mean that the output file will be 8K (unlike the F65 which produces 4K from an 8K sensor). With that kind of pixel count, it almost won't matter (for cinematic purposes) whether you do a frame extraction or use the whole sensor width, as either choice will involve scaling down.

 

[Marc Wielage]

All things being the same, conventional wisdom tells us that the high-res file scaled to a lower-res format will always trump a recording in said lower-res format.

That's correct. We knew 10 years ago that if you took a 4K scan of a 35mm negative and downrezzed it (even on the fly) to 2K for the D.I., it wound up looking better than a straight 2K scan. Having seen direct comparisons of 2K scans and 4K scans (and 4K -> 2K scans), I can tell you that the biggest difference was the first two. The 4K scan is better but it's only a skosh better than the 2K scan downrezzed from 4K. Certain kinds of elements and backgrounds will bring out the details, but in general, it's not a gigantic and unsubtle difference. Once you uprez a decent 2K scan back up to 4K, I'd say about half of the differences vanish -- depending on the methods used and if any sharpening or grain reduction is applied.

These tests were done with John Bailey, ASC and several other DPs at Kodak's Cinesite in the early 2000s. It was quite an education to see why and how things changed in the translation. Digital capture and digital projection adds a different element, but the same principles still apply.

I just thought of something that I missed before. I said earlier that a 33% loss of resolution - yes, it's 33% - seemed too much. However, the 5K frame is 2:1, not 2.4:1. To crop the 5K frame to get 2.4:1 you have to lose about 427 (2560-2133) lines, or a little over 16% of your resolution.

I think you guys are getting way too wrapped up with numbers. Bear in mind that Fincher made the decision that he needed the extra room in order to change his mind about framing later on in post (an idea I agree with 100%), and this flexibility trumps everything else. If having the ability to pan left 10% or tilt down half an inch tells the story better and/or saves the shot, then the resolution is the least important factor here.

The only time I really noticed a major difference between 2K and 4K was in bright, perfectly-focused, steady shots with a lot of vertical lines, particularly tall buildings with a lot of glass, cityscapes, stuff like that. Then, we picked the 4K from 2K 10 out of 10 times. The rest of the time, the comparisons were much, much harder to do than you might think -- even by experienced VFX people and cinematographers.

 

[paulherrin]

many lenses offer better edge to edge sharpness wide open at the 4k crop, but introduce more falloff on the 5k image. if the falloff is not desirable, the 4k extraction could be sharper on the edges than scaling the 5k image. those and other factors aside though, a 5k generic 5k image scaled down to 4k should display sharper than a 4k image (scaling method must be taken into account, however). there are a lot of benefits, as michael points out, to shooting with some look around for reframing/etc.

 

[Martin Stevens]

many lenses offer better edge to edge sharpness wide open at the 4k crop, but introduce more falloff on the 5k image. if the falloff is not desirable, the 4k extraction could be sharper on the edges than scaling the 5k image. those and other factors aside though, a 5k generic 5k image scaled down to 4k should display sharper than a 4k image (scaling method must be taken into account, however). there are a lot of benefits, as michael points out, to shooting with some look around for reframing/etc.

Very nice post Paul.

Thanks.