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 :)

Moire with a Scarlet ??

George D.

George D.

Well-known member
Joined
Oct 22, 2008
Messages
980
Reaction score
0
Points
0
Location
N. California, USA
Will there be an issue, any issue, with moire patterns showing up on Scarlet footage?

It is a serious concern because we shoot a lot of macro of insects. Since insects have large, compound eyes with hundreds of honey-comb like facets covering them, if that should create moire patterns, it would ruin the shots. Diffusing the image in post, is NOT an option for us.

Cameras with CMOS shutters are considered notorious for this, and I just want to know if it COULD BE an issue with the upcoming, interchangeable lens Scarlet.

If it could be a problem with the Scarlet, how about the RED-One ??

Thanks.
 
As Graeme just said in the other thread on this, RED knows the value of proper low-pass filtering to avoid this very issue.

Graeme in particular is a strong champion of avoiding any aliasing in images... I doubt he'd ever allow any of the cams, Scarlet included, out the door without putting it thru an aliasing torture test to make sure you get pristine images even under the difficult subject matter you describe.

-sc
 
Exactly! While testing we shoot zone plates that exercise all spatial frequencies to ensure that aliasing is negligible.

Graeme
 
We're shooting with the new Mysterium x Chip and we have a lot of trouble with some kind of moire effects. Thin lines on clothes or roofs let the camera freak out and make not only moire. It also produce strange colors in this area in some frames.
Did someone have seen this before?

Kind regards

Tim
 
Tim, we have never experienced that on RED. Please arrange to send me the R3D file immediately.

Graeme
 
Lens and f/ stop

Lens and f/ stop

I think I read someone else note that the OLPF on the MX or the new sharper de-Bayer seem to show some artifacts that the older sensor board, de-Bayer, and OLPF did not right after MX started shipping a while back.

There is the issue while testing of the impact of the lens on the OLPF since the thickness of the OLPF being about 2mm can cause enough negative spherical aberration to maybe be visable depending on the f/ stop and make the OLPF look like it is working better than it is.

If you adjust the de-Bayer to reduce the sharpen of close green pixels you may tone down the artifacts. But a stronger OLPF is a better way to go since its better to have a soft 3K image than a rough 3.5K image.

I know they have been trying to get closer to 4K because people say that its not a 4K camera in the sence of a film scan made with a monochrome 4K camera, but that does not matter since you cannot see 4K on a 35mm film print projected other than some less aliasing, that is if the source image does not show aliasing itself.

Anyway, getting back to the lens and f/ stop, the test charts should be shot with enough daylight balance light that the lens is at prime stop, and then reduce the brigtness to double check what happens at larger stops and a little out of focus. And then work the other way to check for a smaller angle ray bundle through the OLPF and micro-lens array.

Try shooting the tests through a magenta filter since the green sharpness may mask some of the artifacts on a monochrome test chart.

Try to incline the camera at about 5 to 10 degrees some one side of the test chart is closer than the other, and rack the focus over the test chart, that way a vertical line will be in focus, the set the camera at "Dutch angles" and try again...

Also you are using a chroma kill in the de-Bayer for such artifacts on a monochrome test chart, but when the color balance is off center the chroma kill may not have its factors as optimized?

If you get a black tray and fill it with small metal ball bearings (you can get them at some hardware stores) and point a single point light source at that, it makes a good test for the OLPF, the small bright reflection of the ball bearings tends to make a bunch of colored spots when you sharpen the image and boost the saturation if the OLPF is not heavy enough. If the OLPF is working you just see white spots. Do not overexpose the ball bearings since that will spread the light and reduce the moire. If you have a zoom lens you can zoom in and out on the ball bearings to see if that has an impact on the moire. And change the f/ stop and lighting, like indoors or the sun's reflecton in the ball bearings. Metal rods side by side can also be another test, since they make high contrast lines rather than dots. I have noticed that stanless steel hand rails shot at a slight angle with sun on them from a distance seem to show aliasing and chroma moire more that just about anything except chrome polish bayonets at a distance with glint on them.

I saw some info on the internet that shows some lenses at some stops can cause interference patterns because of the ring shaped circle of confusion, you see the moire when the lens is a bit out of focus one way but not the other, so not only do you need to vary the light and f/ stop, but also the lens focus a bit in and out to compensate for that effect and the "plate" effect of the OLPF on the point of best focus changing as the f/ stop is changed.

If you focus wide open and stop down, the lens will not be in the plane of best focus, and will show less moire and aliasing.

There could be more of an issue with Scarlet with the "plate" effect since the OLPF is thicker in ratio to the frame height and people will try to use faster lenses maybe so when they stop down without adjusting the de-Bayer the OLPF comp and sharpen may be set too high.

With OLPF its better to span enough radius to cover the red and blue distance, I remember your saying that you have it adjusted mid-way between the shorter green span and the longer red & blue span, so the red and blue are a bit under blured, maybe?

On the subject of OLPF, have you noticed an impact on the red noise using a "green glass" IR cut vs. an interference IR cut built into the OLPF+IR/UV?

It seems the "green glass" type cut more red at about 660-680nm maybe?

What lens and f/ stop was Tim using when he saw the issues, and can he see them in a full size TIF frame output?

Maybe he can enlarge the "bad part" 2x and crop from the TIF to make a JPG to post here (make it 2x to overcome the JPG blur from compression)?

In some of the test images from the Acam dII I could see chroma moire in the center of the images but not at the edges, so the lens may show more issues near the center of the image. I talked to them about that and they were going to look into it, they may not have seen it since in Sweden they are working indoors under lower light levels, not daylight on real world subjects with maybe better lenses, maybe... I think they are going to fix that issue with a OLPF that can span the pixel size better that just the lens blur.

One odd thing about the chroma moire, with some kinds of chroma area filters you do not see the off color spots, rather the area fill converts the chroma moire into a color cast, such as pink highlights.
 
We're not seeing any significant artifacts on M-X, even using pretty brutal test patterns that have stronger high frequency detail than you'd ever see in the natural world shooting.

Graeme
 
Tim is a 1st time poster with a problem that no one has ever seen before on any Red camera. Makes you wonder.
 
Well, given I pm'd him as soon as I saw the post, and posted here asking for the R3D, and given as Phil points out, he's reporting an issue that is unseen in all out testing, and without evidence to the contrary, I can only guess that he was actually using a 5D2, not a Mysterium-X equipped Red One.

Graeme
 
Hey guys,

As we are already on it I have seen some pretty significant aliasing/moire on Red footage - after I transcoded the files to ProRes 422 HQ.

A good example would be panning over moving leaves of a tree, which resulted in quite and ugly effect...

I understand that this might be due to the conversion. But it would be greatly appreciated if one of you pixel gurus could point out a good conversion workflow to HD to avoid/minimize aliasing and moire.

After all most of the Red Footage shot these days is ingloriously down converted to 1080i at one point....
 
axel ebermann said:
Hey guys,

As we are already on it I have seen some pretty significant aliasing/moire on Red footage - after I transcoded the files to ProRes 422 HQ.

A good example would be panning over moving leaves of a tree, which resulted in quite and ugly effect...

I understand that this might be due to the conversion. But it would be greatly appreciated if one of you pixel gurus could point out a good conversion workflow to HD to avoid/minimize aliasing and moire.

After all most of the Red Footage shot these days is ingloriously down converted to 1080i at one point....
Click to expand...

Using a smoother re-size filter can avoid / reduce it.
But I think independently the way you use, the danger begins when this signal is converted to interlaced NTSC/PAL. IMHO.
 
axel ebermann said:
Hey guys,

As we are already on it I have seen some pretty significant aliasing/moire on Red footage - after I transcoded the files to ProRes 422 HQ.

A good example would be panning over moving leaves of a tree, which resulted in quite and ugly effect...

I understand that this might be due to the conversion. But it would be greatly appreciated if one of you pixel gurus could point out a good conversion workflow to HD to avoid/minimize aliasing and moire.

After all most of the Red Footage shot these days is ingloriously down converted to 1080i at one point....
Click to expand...

Can you post an example?
It sounds more like strobing rather than aliasing.
 
axel ebermann said:
Hey guys,

As we are already on it I have seen some pretty significant aliasing/moire on Red footage - after I transcoded the files to ProRes 422 HQ.

A good example would be panning over moving leaves of a tree, which resulted in quite and ugly effect...
Click to expand...

How is that possible? Shouldn't it be so blurred that any kind of aliasing would be impossible? That should be true even on the 5d...unless we're talking about very slow movement...
 
That's why we need to see these things to diagnose / debug properly. Sometimes words are not adequate to describe an issue and a movie or still really helps.

Graeme
 
Downsize method and pre-blur radius

Downsize method and pre-blur radius

Anytime you downsize a digital image, no matter where is comes from you can get aliasing.

If you use "nearest neighbor" resize it also "amplifies" and residual chroma moire in the large image when made smaller, whereas if you average ALL the pixels going downsize it blends out the chroma moire and aliasing, you how you downsize can make the chroma moire worse or better.

I was getting some awful aliasing going from 4k images to smaller sizes like 512 wide for the proxy frames until I made some changes to how I do the quick downsize for that use.

In addition to doing a full all pixel average, I found that if you reduce more that 50% from 4K images if they are too sharp even the all pixel average is not enough, you need to do a half pixel gauss type blur on the 4K image before doing a all pixel resample to sizes like 1920 wide.

It is not possable to have "pixel sharp" digital images for anything other than line H or V graphics since you get "runners" due to the aliasing.

Digital images MUST be about 0.5 to 0.707 the resolution (at high MTF) of their pixel count to not show aliasing, so at most a 4K image should show between 2K and 2.8K at high MTF to not show aliasing on moving high contrast edges.

2K digital images would then show maybe 1414 resolution which is just above the 1280 needed for 35mm film projection, well some film recorders are 1828 wide and called 2K so they give 1292 which is about what has been measured as the value you see in a movie theatre off a 35mm print.

At lower MTF some small details can alias since the alias is under the contrast the eye can see clearly, maybe.

Anyway, the point being that you should look at the 4K TIF files, not some JPG or compressed downsized image that is being resized by god knows what method to display on your monitor, to see what is in the de-Bayered data.

If you are going to downsize a 4K image to HD size or smaller you need to run a half pixel to three pixel radius gauss blur on it before you downsize to avoid aliasing later when the over sharp images are displayed or resized on the viewing device, the edges must all have some softness to them or you see the "runners" when things move etc.

These resize problems are not due to issues in the camera perhaps, but to over sharpening of the full size data or not using a OLPF with a red-blue span to increase the still frame resoultion at the expence of somewhat too sharp results. But even if the OLPF is enough and the de-Bayer is not to sharp, any digital image going downsize needs to be soft before the resample and not over sharp.

What I have noticed with the camera data I am working with is that to get best results for say DVD size results vs. film recorder size results is that the whole de-Bayer settings need to be re-adjusted to optimize the S/N ratio vs. aliasing issues. If you optimize the de-Bayer for the target size you can get quite a bit better results, so making 4K images as the ONLY de-Bayer, then re-sizing those is not the best way to go, the de-Bayer shoud be adjusted for each resize target so you need to de-Bayer many times from the sensor data to get all the needed target sizes required if you want the best results "possible".

With the limits of the programs that read R3D, you can at least make:

1) A sharp set of 4K TIF for use with 4K projection and post. Use medium settings of the OLPF comp and sharpness at mid-range.

2) A soft set of 1920x1080 DPX for post for HD. Use medium settings of the OLPF comp and sharpness at low.

3) A very soft set of SD size frames for DVD use. Use low settings of the OLPF comp and sharpness at minimum.

==

About the test chart shooting, it would be good to have a slow DC motor drive on the tripod head that is used to shoot the test charts so the camera can be panned and tilted with slow motor drives, and limit switches used to reverse the motors, that way the camera and be in constant moton at various speeds while shooting the test charts to be able to look for any slight sign of runners.

The contrast the test charts are looked at has in impact on seeing the aliasing, so you need to boost the contrast at least 8x on the test chart shots and boost the saturation at least 8x as well, because some people will be using LUT that boost the contrast and saturation that much in post, then you can see what they see that would not be visable in the normal "RAW" output since the eye can not see low contrast details that do not move.
 
Images shot on film can alias when shown in HD or SD too -- any fine lines in an image will interact with the grid pattern of a display device. As long as you don't bake-in aliasing into the original, you have the flexibility of softening those fine details so that they do not alias... but at some point, you are going to have to decide if you really want to soften the worst offenders so that there is never any aliasing on any monitor, or just learn to avoid shooting subjects that are prone to aliasing.
 
You must log in or register to reply here.
Back
Top