Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

2K Soft Focus on all Red Ones

Excuse the JPEG - She's 19' 1/8" away from focal plane, RED 18-50 zoom. 2K @ 72fps.
aliceshot.jpg


Perhaps you can't see it here, but she's sharp...
 
Carsten Fenstermacher said:
Ah. It looked like there was one but it looked to be not very strong.
How strong or weak is the new one?
Click to expand...

Just right?:laugh:

The OLPF in the RED is pretty aggressive. You certainly won't get any aliasing but it does soften the image a bit. The filter in the Phantom is tuned just a touch beyond the frequency point of the sensor. At 12.5 microns, it can start to get to be a fairly thick chunk of crystal if one goes too far.
 
Goldilocks and three OLPFs :)

I've found that Red low pass is quite pleasing as it doesn't make the talent look years older. I've not shot older people on phantom to know how it compares but it would be interesting to see.
 
keicol said:
The point is that we have a great 4K camera, why can't it shoot 2K with the expected drop in resolution and the expected effects from going to a wider lens and maybe a bit more open iris but still give us a usable picture with subjects at about 15'?
Click to expand...

I guess I'm not sure just what you expect to see that you're not getting. Nominal resolution of the RED One falls between 3.2K and 3.6K from a 4K frame. From a 2K crop of that frame, the same still applies -- 1.6K to 1.8K nominal measured resolution. This is what I've seen when I've shot a res chart in 2K as well. Unlike the F900, you have to process and add your sharpening as you develop in post.

And to everyone who keeps saying how superior these other HD cameras are compared to R1 2K, go point them at a resolution chart. EX1, EX3, F900, HPX3000, etc.. none of them deliver a full 1920x1080 measured resolution.

If I have to guess at what's going on here, and that's all I can do anyway since we're not getting any real info here, I would say there's a breakdown in the post process somewhere.

How are the 2K images being compared to the 4K images. You say the 2K is being shot off-speed (of course it is, why else would you be shooting 2K, right). How do the exposure settings differ from the 4K image? How are the images being processed prior to going to the FX guys?
 
Isn't it a simple matter of 2K photosites actually being equivalent to 1.6k pixels? Ergo, you have to uprez to get the 2K (1.6k) to fit in a 1920x1080 window, or 2048 x whatever window, for the DI. The 4K footage, more like 3.2K is more than enough information to downrez to the same 2K frame. The uprezzed "2k" originated footage will have to be soft compared to the downrezzed 4k footage.

If you were downrezzing both of them to a 720P finish, perhaps they would be more comparable. But the uprez/downrez matching seems to be where the problems lies. Correct me if I'm wrong.
 
I really do not know what people here talking about?

Soft focus!!!???

But that's a really anti-propaganda to RED1...

And here we go... shot at March 14, 2008 with Zeiss Arriflex 85mm T2.1 Standard on RED1 (slow-motion)...
...

varispeed_2K_shakePRHQ.jpg

Shot at March 14, 2008 with Zeiss Arriflex 85mm T2.1 Standard on RED1 (slow-motion 112 FPS)...

2K slow motion test footage 112 FPS shot on RED1>>>
 
Joe G. said:
Isn't it a simple matter of 2K photosites actually being equivalent to 1.6k pixels? Ergo, you have to uprez to get the 2K (1.6k) to fit in a 1920x1080 window, or 2048 x whatever window, for the DI. The 4K footage, more like 3.2K is more than enough information to downrez to the same 2K frame. The uprezzed "2k" originated footage will have to be soft compared to the downrezzed 4k footage.

If you were downrezzing both of them to a 720P finish, perhaps they would be more comparable. But the uprez/downrez matching seems to be where the problems lies. Correct me if I'm wrong.
Click to expand...

A debayered 2k image is still 2048x1152 (16:9) pixels, each fully defined with RGB color data. The optical resolution as reproduced by a test chart will be lower. The same is true of an HD 1080p 3 chip camera, 750 TV lines being close to the maximum vertical resolution one will see on a test chart with the best high end cameras and top of the line lenses.
 
"A debayered 2k image [WITH THOSE SETTINGS] is still 2048x1152 (16:9) pixels, each fully defined with RGB color data. The optical resolution as reproduced by a test chart will be lower."

Then, it's lower. It's been scaled up by the algorithm.

You can have Photoshop turn a 300x200 pixel image into a 3,000 x 2,000 pixel image. So what?

And yes, Sanjin, your test is a bit soft.
 
Joe G. said:
"A debayered 2k image [WITH THOSE SETTINGS] is still 2048x1152 (16:9) pixels, each fully defined with RGB color data. The optical resolution as reproduced by a test chart will be lower."

Then, it's lower. It's been scaled up by the algorithm.

You can have Photoshop turn a 300x200 pixel image into a 3,000 x 2,000 pixel image. So what?

And yes, Sanjin, your test is a bit soft.
Click to expand...

It's not "scaled up". If you shoot 4K RAW and the image measures 3K on a line resolution chart, it doesn't mean that 4K RAW was debayered into 3K RGB files and then was scaled back up to 4K RGB. It's just the difference between the size of the files versus the real-world resolution in the actual image. Just because an image is soft doesn't mean it's because it's a rescaled smaller file. I could shoot on a 16K camera with a crappy lens and diffusion and get less than 1K of measurable resolution, but that doesn't mean I'm actually taking 1K files and then scaling those files up to 16K.
 
Wow, this thread has seen some action. It has been a long week for me already, but I was able to shoot some more quick and dirty tests to see what is going on here. I haven't had a chance to critically review these yet, but if you want to download the TIFF's you can do so here:

2k Soft Focus Test: (301MB)
http://www.ryanewalters.com/downloads/Res-TIFFS.zip

All footage was shot using the Red 18 - 50mm in 2:1 Redcode 36, debayered at maximum using Red Alert. Three images were shot at three distances. A 2k image at 18mm, the 4k at that same length, and then a 4k at 38mm which is the same FoV as the 18mm. I also took the 4k image into photoshop and cropped out just the center to get a direct comparison between the 4k cropped and the full 2k.

Focus was set and locked for all tests, the only change was the resolution and / or the focal length. Focus was set on the chart in the center of the frame. (Yes I know the chart is upside down. :) ). The distances used were 6', 16', and 37'. Focus was checked with a tape measure, the lens readout, and it was visually verified using the pixel for pixel function on the LCD.

Hopefully this is helpful. As I said, I haven't critically looked at these yet. I'm off to bed as it has been a long day. :)
 
Were all those shot wide open? IMHO, 5.6 would have been more useful.

Comparing the 2k to a crop from 4k it looks the same, ya?
4k 38mm scaled to 2k looks sharper as expected than 2k but is also adding lens factors as the 18mm will perform differently than the 38mm part of the lens.
 
"If you shoot 4K RAW and the image measures 3K on a line resolution chart, it doesn't mean that 4K RAW was debayered into 3K RGB files and then was scaled back up to 4K RGB"

Not exactly what I was saying. If there is only enough information to define 3K accurately, yet the algorithm is designed to turn it into 4K, then there is some sleight of hand involved.

I believe this has something to do with the chroma vs. luma data sets. I'd like a fuller explanation, too, of the "78%" number, and what exactly it means (from the company).
 
The algorithm doesn't know how sharp the original photography is -- it is transforming a monochrome image which was shot with a color filter mosaic pattern in front of it, and turning it into a color image in whatever file size you want. It's not somehow going from 4K to 3K and back to 4K again because that presupposes it knows that the image was 3K in measurable resolution. It's taking 2000 green photosites and 1000 each red and blue photosites and making some sophisticated guesses as to the color content of the surrounding photosites and then reconstructing a 4K RGB image. The fact that the image itself only measures out as 3K in line resolution is a byproduct of the whole system. It's not constructing a 3K RGB file that is then upscaled to 4K RGB. Like I said, it's a transformative process.

You just have to stop confusing the amount of resolution you can see on the screen with the size of the files or the number of photosites in the sensor. The Optical Low Pass Filter alone would account for the fact that you can't get an exact number of lines in the picture to match the pixel grid -- and because of aliasing, you probably wouldn't want that anyway.

Some sleight of hand involved? The whole process is a sleight of hand! You are converting a monochrome image into a color one using a limited number of clues as to the true color at each photosite.

There's some interesting stuff here to read:
http://en.wikipedia.org/wiki/Demosaicing
 
Last edited:
Ryan E. Walters said:
Wow, this thread has seen some action. It has been a long week for me already, but I was able to shoot some more quick and dirty tests to see what is going on here. I haven't had a chance to critically review these yet, but if you want to download the TIFF's you can do so here:

2k Soft Focus Test: (301MB)
http://www.ryanewalters.com/downloads/Res-TIFFS.zip
Click to expand...

OK, I took the bait and downloaded them... So far, just looked at the 18mm 6ft tiffs, the 2K crop from 4K and the 2K shot look essentially identical. Of course, the two shots don't even line up the same and the exposure is wack as well as the lighting. What color temp are we dealing with here? About 2900K?

I took those two tiffs into Photoshop at 16bit and corrected levels and picked my whitepoint and essentially got the two images to match by eye. Then I nudged one over the top of the other, they're nearly a perfect match. Placing the top image in difference mode, I only see faint and rather sharp outlines of most high contrast edges, which is to be expected due to the shifted framing between the two images and sensor noise.

Sharpening with a high pass and unsharp mask works fairly well considering the under-exposure.

What aperture were these shot at? The images seem to feel as if they were shot full open at T3 as flare and CA seem a bit excessive for the 18-50, but it's hard to judge based on these lighting conditions and lack of info.

If a wide open aperture is indeed what we're seeing here, then I'm not really seeing a problem. And no discernable difference between a 2K shot and a 2K crop of a 4K shot.

How were the images processed other than full debayer? What color space, white point, OLPF compensation, ISO level, etc..?
 
Can we just get Graeme or other Red rep to say if there is any difference or not and close this thread. My little head can hardly get through all these advanced posts and I would love to get to the bottom of this... something tells me this will end up like the "stuttering while panning on Red" thread... it just seems bogus. Aside from a little lower effectiveness of the compression methods what the hell could be going array here?

Pls someone close this down with a definitive answer.


:emote_headwall:
 
Jeff Kilgroe said:
the 2K crop from 4K and the 2K shot look essentially identical.
Click to expand...

It seems obvious to my naked eye that the 2K shots are soft. Are we looking at the same samples? I haven't done any Photoshop test but just by looking at them it seems obvious to me that the wine bottle graphics are very soft in the 2k version compared to the 4k version in Ryan's test.
 
You must log in or register to reply here.
Back
Top