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 :)

Noise

Jeremiah McLamb

Jeremiah McLamb

Well-known member
Joined
Apr 18, 2007
Messages
206
Reaction score
0
Points
16
Age
42
Website
www.jerfilm.com
I feel like I'm seeing way too much noise in my viewfinder and on the r3d files on the computer. Sometimes the evf is so bad its practically unusable. Here are some sample 2k images from some 4k source r3d files...all 320 iso

It may just be me and I'm just expecting way to much from RED...but the shadows are just way noisier than I would like them to be...and the one image with the blue noise is obviously not good.

Also, all of my exports out of redcine, rushes, or alert are all real noisy as well. I've only rendered out to proress hq...but to me its too noisy to finish on like some people are saying they're doing.

www.jerfilm.com/image01.tiff
www.jerfilm.com/image02.tiff
www.jerfilm.com/image03.tiff

thanks
jeremiah
 
Jeremiah, that does look seriously not-right. If you're in RedAlert, can you post a screengrab of the shot and settings?

Graeme
 
I had some stuff that had this look yesterday when I mistakenly hit the auto WB button when the camera wasn't pointing at a white card. Try resetting the white balance to the tungsten or daylight preset and see what happens.
Also do a Black balance.
Your images look nothing like normal.
Good luck,
Shane.
 
ok...i changed the color temp in red alert and the blue noise went away in the first pic. But is that what's going to happen all the time if I choose a low kelvin temp?

Also, what about the other pics? Is that much noise in the shadows normal? I'll do a reinstall of build 16 today and a fresh black balance and let you know what that does..

thanks
jeremiah
 
jerfilm said:
ok...i changed the color temp in red alert and the blue noise went away in the first pic. But is that what's going to happen all the time if I choose a low kelvin temp?

Also, what about the other pics? Is that much noise in the shadows normal? I'll do a reinstall of build 16 today and a fresh black balance and let you know what that does..

thanks
jeremiah
Click to expand...

It depends on how much exposure the blue channel got.
For a daylight balanced camera, blue is less sensitive than red and green.
When you have very warm lights, you adding to the underexposure because there's very very little blue light. So, when you try go balance the image back to daylight, you have to gain up blue alot because in effect you've exposed the blue channel a couple of stops under the other channels.
 
I invite you to see a thread I started long time ago.
I still dont know how the camera behaves in some low light situations in which you cant always control the amount of light there is available. So far, my tests have induced that either you overexpose the image (so there is more detail in the shadowed areas) and then bring the dark areas back down in RA/RC with the curves, or expose insanely correct and over lit the shadow areas in order to process in the same way afterwords. For some, this is wrong because there is more usable data in the low f-stop range than in the highlight headroom, but this is my personal opinion and this is what I do lately. I have read, that the camera doesnt crush the black level by itself. The cam doesnt know whats black or not, thats for you to decide, so there may be some more information than the one you wanted because it doesnt interpret the black level as you do, so you have to tell the cam whats black, NOT with exposure necessarily, but with the contrast and curve tool in software.
Another thing I do, is always rate the camera between 4500 and 6000 degrees kelvin. And color correct in RA/RC. It gives me still the best results in those kind of situations.
Here is the link:
http://www.reduser.net/forum/showthread.php?t=16288
And then again, I am still looking for a definite answer on this matter aswell.
My 2C
 
estebanred said:
I invite you to see a thread I started long time ago.
I still dont know how the camera behaves in some low light situations in which you cant always control the amount of light there is available. So far, my tests have induced that either you overexpose the image (so there is more detail in the shadowed areas) and then bring the dark areas back down in RA/RC with the curves, or expose insanely correct and over lit the shadow areas in order to process in the same way afterwords. For some, this is wrong because there is more usable data in the low f-stop range than in the highlight headroom, but this is my personal opinion and this is what I do lately. I have read, that the camera doesnt crush the black level by itself. The cam doesnt know whats black or not, thats for you to decide, so there may be some more information than the one you wanted because it doesnt interpret the black level as you do, so you have to tell the cam whats black, NOT with exposure necessarily, but with the contrast and curve tool in software.
Another thing I do, is always rate the camera between 4500 and 6000 degrees kelvin. And color correct in RA/RC. It gives me still the best results in those kind of situations.
Here is the link:
http://www.reduser.net/forum/showthread.php?t=16288
And then again, I am still looking for a definite answer on this matter aswell.
My 2C
Click to expand...

Can you rephrase the question here?
 
I would but I feel intimidated by your avatar. :)
Now seriously... :P
My question would be the same as posted in the other thread:
How do you avoid noise in the shadowed areas?
Not nice looking noise, but crazy noise. Noise that draws your attention too much and you cant bare. In other words I guess after all I have read and tested the question would be how to get nice rich blacks instead of noise?
Thanks!

EDIT: I think maybe the thing with rich blacks I can easily get by adding curves and contrast to the image. But the gradation from correctly exposed areas towards the blacks is what I find is giving me a lot of trouble with the noise.
 
estebanred said:
I would but I feel intimidated by your avatar. :)
Now seriously... :P
My question would be the same as posted in the other thread:
How do you avoid noise in the shadowed areas?
Not nice looking noise, but crazy noise. Noise that draws your attention too much and you cant bare. In other words I guess after all I have read and tested the question would be how to get nice rich blacks instead of noise?
Thanks!

EDIT: I think maybe the thing with rich blacks I can easily get by adding curves and contrast to the image. But the gradation from correctly exposed areas towards the blacks is what I find is giving me a lot of trouble with the noise.
Click to expand...

Because we preserve the noise floor in processing, it is essential to put a toe and pull down the blacks in grading. Of course, it's all relative to scene contrast, exposure and how much highlight vs shadow range you prefer. There's no right answer because everyone has different tolerances for where to place the range.
 
Thanks Deanan, I totally agree.

BUT

There is a fact, the camera behaves very differently when you shoot tungsten or daylight. I feel the camera is almost unusable when shooting in 3200K because the blue channel is so crappy that it ruins the whole image. Not the case in daylight or when scenes are lit with HMI. The camera works flawlessly. Those images you can CC in post, images produced under tungsten lighting IMHO need a lot of post work and noise reduction. I will do a test and post it here during the weekend.
Thanks again, we really appreciate this kind of feedback.
 
estebanred said:
Thanks Deanan, I totally agree.

BUT

There is a fact, the camera behaves very differently when you shoot tungsten or daylight. I feel the camera is almost unusable when shooting in 3200K because the blue channel is so crappy that it ruins the whole image. Not the case in daylight or when scenes are lit with HMI. The camera works flawlessly. Those images you can CC in post, images produced under tungsten lighting IMHO need a lot of post work and noise reduction. I will do a test and post it here during the weekend.
Thanks again, we really appreciate this kind of feedback.
Click to expand...


It does depend on what you do through WB and color correction and where you expose it. Whenever you take any daylight balanced media, film or digital and shoot it tungsten you're under exposing the blue channel and then pulling it back up to balance it back to daylight (or balance it optically).

The difference is that the camera is behaving differently in different light (because it doesn't) but in how you move the colors to another temperature. By shooting with the intention of correcting 3200k back to 5600k you're starting out with much less blue light and gaining up alot more than normal to get it to look like daylight.
 
I agree. The thing is that if you balance your WB on camera you are playing with the GAIN, arent you? And that brings up noise. When you try that and set it in 3200K and light with tungsten it should look right. But it doesnt, it looks noisy. Its better to light with tungsten and set the WB to a higher K degree and then move the gains in POST which wont bring that much noise. I tried to correct optically once with a 80A filter, but the result wasnt that amazing. Still looking for the best way to have a tungsten shoot and make it look right.
 
estebanred said:
I agree. The thing is that if you balance your WB on camera you are playing with the GAIN, arent you? And that brings up noise. When you try that and set it in 3200K and light with tungsten it should look right. But it doesnt, it looks noisy. Its better to light with tungsten and set the WB to a higher K degree and then move the gains in POST which wont bring that much noise. I tried to correct optically once with a 80A filter, but the result wasnt that amazing. Still looking for the best way to have a tungsten shoot and make it look right.
Click to expand...

WB in camera is only metadata and does not affect the image recorded at all. The only user settings that can affect the image are shutter speed and lens aperture.

Optically filtering will bring you close to shooting daylight.
 
Gregor Hagey came to an interesting conclusion, which you can in the thread below:
http://www.reduser.net/forum/showthread.php?t=18042


The short answer is develop your footage at high K temp, say 4500-6000, as appropriate, and then CC in CC software to obtain the right look.

This makes sense, as increasing the blue gain will make the image look right in terms of K temp, but will increase blue channel noise.

(This is supposition on my part, so take it with a grain of salt) Using an HSL correction (like the FCP 3-way) is using information from the other channels to create a new blue channel, rather than simply gaining up blue. So you bypass this problem... (End supposition)
 
AngusChandler said:
Gregor Hagey came to an interesting conclusion, which you can in the thread below:
http://www.reduser.net/forum/showthread.php?t=18042


The short answer is develop your footage at high K temp, say 4500-6000, as appropriate, and then CC in CC software to obtain the right look.

This makes sense, as increasing the blue gain will make the image look right in terms of K temp, but will increase blue channel noise.

(This is supposition on my part, so take it with a grain of salt) Using an HSL correction (like the FCP 3-way) is using information from the other channels to create a new blue channel, rather than simply gaining up blue. So you bypass this problem... (End supposition)
Click to expand...

I personally like this technique also. It's basically the equivalent of desaturating only the shadows a bit.
 
Deanan said:
WB in camera is only metadata and does not affect the image recorded at all. The only user settings that can affect the image are shutter speed and lens aperture.

Optically filtering will bring you close to shooting daylight.
Click to expand...

I have made a few test shoots in which I set the camera WB differently and then review the results in post. IMHO the results are different whether you set the WB closer or further the 5000~K mark. So I thought the RAW file was exposed differently. So what you are saying is really shocking news to me. I had read before about it but thought it couldnt be because whenever I shoot at 3200 my images look terrible and I cant work with them even in post. So I always set it in 4000 when shooting tungsten and then CC in RA. David Mullen stated in that thread that using the 80's filter wouldnt be much of a solution because it eats a lot of light, so optically filtering is really not an option unless you have a 1.3 lens. Not the case. Moreover, the example technique that was posted brings up an aesthetics issue that has to be on your mind whenever you judge this kind of tests and this is that after looking the graded scene you can see that the look that they wanted was yellowish, so setting up the camera in the 4000 mark and then developing on 5600 would bring up the orange and lower the blue avoiding the noise in that channel. BUT if the look you were after was not yelowish but more neutral, the blue channel wouldnt be that lowered and the noise could still be there, so for me the test works in a way and then again in another way it doesnt. WHY? Because when you shoot in tungsten and want it rated neutral, not tinted, the noise in the "underexposied blue channel" is unbearable. Tonight I will make a test of a scene that has been problematic and I will post it for you to share your opinions and bring up feedback.
Have a good weekend!
 
Colortran...

Colortran...

estebanred said:
I have made a few test shoots in which I set the camera WB differently and then review the results in post. IMHO the results are different whether you set the WB closer or further the 5000~K mark. So I thought the RAW file was exposed differently. So what you are saying is really shocking news to me. I had read before about it but thought it couldnt be because whenever I shoot at 3200 my images look terrible and I cant work with them even in post. So I always set it in 4000 when shooting tungsten and then CC in RA. David Mullen stated in that thread that using the 80's filter wouldnt be much of a solution because it eats a lot of light, so optically filtering is really not an option unless you have a 1.3 lens. Not the case. Moreover, the example technique that was posted brings up an aesthetics issue that has to be on your mind whenever you judge this kind of tests and this is that after looking the graded scene you can see that the look that they wanted was yellowish, so setting up the camera in the 4000 mark and then developing on 5600 would bring up the orange and lower the blue avoiding the noise in that channel. BUT if the look you were after was not yelowish but more neutral, the blue channel wouldnt be that lowered and the noise could still be there, so for me the test works in a way and then again in another way it doesnt. WHY? Because when you shoot in tungsten and want it rated neutral, not tinted, the noise in the "underexposied blue channel" is unbearable. Tonight I will make a test of a scene that has been problematic and I will post it for you to share your opinions and bring up feedback.
Have a good weekend!
Click to expand...

You can get more blue light out of Practicle lights by using a "ColorTran" or other transformer to raise the voltage this boosts the blue light by up to 400% or more, see page 287 in American Cinematographer Manual 1st Edition.

Quote:

186volts (2900K120vlamp) 3400K
violet+650%
blue+412%
yellow+275%
red+210%

so you get two stops more blue and one stop more red.

that should give you about 12db better S/N for the sensor since the "sensor" gain will be lower for the blue and green.

Because the images are compressed before de-Bayer having the light not right on the balance point can make more noise due to compression artifacts, if you looked at the data going into the compression the blue image is dark, and so would undergo a higher compression ratio than the red or green images because the compression reduces data in the dark areas, the best way to deal with that issue is to use blue filters and boost the lights so the light going into the camera is always balanced.

The "ColorTran" was developed 1939 by L.V. Grover of LA, we have several of them and use them to get more light and less blue noise on film which also has the problem of higher noise in the blue. In my software I have a adaptive filter that lets me filter the red-green-blue bandwidth to reduce blue noise in film scans, it would also work on digital images, you can see an example in my post about the NoNoBe frame, although that is not a super good example since the footage was soft, it is better if the footage is in focus so the filter can tell detail from noise, if the shot has high noise and is out of focus there is less you can do to save it...
 
You must log in or register to reply here.
Back
Top