Optical Filtration vs. Digital Filtration

[Mike Nagel]

Hi,

since there is a lot of discussion lately in the Dragon threads and some other threads regarding if optical filtration can be truthfully imitated in post, I thought it'll be nice to start a separate thread and maybe share experiences or examples here...

So what do you prefer / use and why ? Optical or digital filtration ?

- Mike

 

[Björn Benckert]

Some filters are impossible to match in post. Try to mimmic a polarizer...

Also when you shoot in odd light conditions you risk starving the camera in any of the 3 color channels. For example we are about to go to SA in the end of the month and shoot on a quite deep reef. Those images needs to have a natural / shallow clear water look with lots of colors. As the reef is deep the blue channel is high and green is low and red is almost non existing. So then to fit all channels within the cameras range we need a CTO85 to hold back blue and green, then as the light levels there are relatively high we can just open up and get proper exposure in all three channels, that gives us a daylight / 5600k look even though we shoot deep down. Something that is impossible to correct in post as we would have close to no signal in the red channel without clipping the blue.

Same but opposite when shooting tungsten, then you can filiter out the red channel and part of the green and get a proper exposure in all channels. Something that is pretty much a need when shooting blue or green screen with tungsten. If you don not filter such situation then you risk having a staved blue channel and if you key on blue you will experience noisy keys and if you key on green well then your picture are left with a poorly feed blue channel and red. And that gives your image a faded color spectrum, as red + a little bit of noisy blue does not really look that smashing... Again this is nothing that you can compensate for in post.

Then there are star filters, flare filters and many optical VFX filters that just can not be mimicked in post.

Another example is lowcon filters that you can use to expand the DR and avoid clipping and underexposure.

And many many more, I do not use much filters but I work a lot with some DP's that does not shot without having a whole stack in the mattebox and from treating their images I can tell you that I would not be able to get even close to the result with whats possible in post.

So to me it's pretty clear that filters are needed for a lot of things and if you are looking for a certain style or need balancing or such and you know you can reach by adding glass in front of the lens then thats the way to go.

 

[Mike Nagel]

Bjoern,

do you feel that you can truly re-create in post what an optical low con filter does to the light ?

this is in regards to the discussion of matching Alexa, blooming highlights, halation and nice highlight rolloff... I understand that in post you'd have much more control, as in what part of the image is affected, but can one truly re-create the effect WITHOUT painting into the image or adding false data, meaning not adjusting / modifying / tweaking existing raw data but adding new, false data to the image... ?

Thanks !

- Mike

 

[Björn Benckert]

Bjoern,

do you feel that you can truly re-create in post what an optical low con filter does to the light ?

this is in regards to the discussion of matching Alexa, blooming highlights, halation and nice highlight rolloff... I understand that in post you'd have much more control, as in what part of the image is affected, but can one truly re-create the effect WITHOUT painting into the image or adding false data, meaning not adjusting / modifying / tweaking existing raw data but adding new, false data to the image... ?


Thanks !

- Mike

I try to explain,

With a lowcon filter in camera you can avoid clipping a bit further. basically you gain DR and loose some small frequency differences. So if you shoot with the filter and by doing so preserve some details in the almost whites those will not be recoverable if shouting without it and clipping them to white.

Do not know if the above made things more clear,bAnother way to look at it:


I find the epic and the MX sensor to be more linear than the alexa that is more low contrast and rounds off more. To me it's arguably better to have a camera like the reds that is more natural / linear and that can be filtered or post processed for lower or higher contrast if wanted. As the "built in low con" of the alexa is far from something that always makes a better picture, it's has a lot of downsides to shoot low con for many situations.

Also as the "normal alexa" gives you a lowcon 10bit log image you do not have to many bits left when you actually graded the image it's more of an 8 bit picture. where as the 16bit log image of the epic kind of leave you with 12bit.

 

[Dave Blackham]

This issue of mimicing Alexa is something that seems to be of interest above other 'looks'. Most of the time for our work we just don't want to but there are times when the look is desirable so Im interested in options to achieve this. It could be interesting to see if it is possible to emulate highlight halation and the effect of a 1/4 low con in post which is whats said to do the trick by in camera filtration.

Im fairly sure you will have a hard time to try get an Alexa to match an Epic though.

 

[David Kuhn]

Any filter that relies on information that the camera doesn't capture can't be duplicated in post. Polarization state is one, as would be a filter that acts on data that is confounded with other data (IR or UV contamination, or if for some reason you wanted to filter a specific part of the visible spectrum). Getting an Alexa look with a Red camera would depend on exactly what is being done to generate it. Things like blooming highlights, rolloff, and halation could likely be recreated (since it's intensity driven and is actually removing information not adding it), but there are a lot of things that can't be fixed in post. For the low contrast filter I'd ask what you'd be using it for...if it's allowing you to extend your dynamic range by reducing highlight levels then it may be tough to duplicate (similar camera setup without the filter will have saturated highlights... data lost).

 

[Les Dittert]

Lowcon filters can lift the blacks up into a range that the sensor can 'see' and resolve details in better. This can not be synthetically done to the non filtered version because the blacks are gone, so nothing to resurrect.

 

[David Kuhn]

Lowcon filters can lift the blacks up into a range that the sensor can 'see' and resolve details in better. This can not be synthetically done to the non filtered version because the blacks are gone, so nothing to resurrect.

With the Epic's dynamic range do you guys often run into situations where you STILL don't have enough and would look at a compression technique like an LC filter... or is it a pretty niche thing?

 

[Mike Nagel]

I knew I should have not mentioned the "matching Alexa" bit, it took attention away from my actual question... ;-)))

Les makes a very good point that the additional information from a low con filter in the blacks cannot be done in post.

but I'm interested - strictly regarding the highlights - if one could truly re-create that "blooming", "halation" effect in post (without adding false information) matching what an optical low con filter does to the highlights ?

 

[Rudi Herbert]

Bjorn,

As usual, you are correct, but I'd like to point out a couple of things regarding the underwater part. I, too, have pointed out many a time that when shooting underwater, it is imperative to use optical filters to balance the light as much as possible before it hits the sensor, since that unbalance cannot be recovered later just cause the image "is RAW". However, an 085 filter, or any other warming filter, is not the best option for this task. There are filters made specifically for underwater work that have formulations much more suited for the actual conditions there, i.e.: with a spectral response that filters out much more blue/green than a warming filter. There are formulations for blue water and green water, where the filter repels either more blue or green, something an 85 does not do. I strongly recommend you use a purpose UW filter over an 85. Best choices are:

UR pro: http://www.urprofilters.com/ Crappy website, great filters, made of optical glass, mounted at the front of your lens, most thread sizes available. John Ellerbrock from housing manufacturer Gates can hook you up www.Gateshousings.com

Magic filter: http://www.magic-filters.com/ Again, crappy website but good filters. Gel strips can fit right over the OLPF. Found no many stores across Europe and US.

Trust me, either version will do MUCH better than an 085, but you can, of course, use whatever you like

 

[David Mullen ASC]

As I mentioned in the other thread, the trouble with post diffusion to create halation is that most post houses opt for the method of blurring a luminance key of the highlights and then layering this back over the clean shot. The problem with this layering technique is that you increase the contrast of the highlights and if you lower them back down again (as you have control over this) you also lower the halation effect. And there is some loss of sharpness in the highlights. It's a pretty look, and flattering for skin tones, but I feel that if you just want a bit of glow around highlights with minimal loss of sharpness and no change in the gamma of the highlights, a light camera filter is better, though keep in mind that designs like Classic Softs and Hollywood Black Magics have a pattern designed to soften detail as opposed to just mist particles like a ProMist or Frost filter might have. Black Magics are a combination of HD Classic Softs and an 1/8 Black Frost, and both increase halation but the Classic Soft portion is designed to soften detail more. Of course, mist filters also soften the image a bit too, just not as much as other types of diffusion.

As for using low-cons, ultra-cons, etc. for increased dynamic range, the theory is that the filter before it hits the sensor (and before recording, compression, etc.) would lift the blacks and some low-level detail along with it up the chain a bit, higher above the noise floor. However, as with techniques like flashing in film, I've found that there is only a 1/3-stop increase in usable detail, most of the perceived increase in shadow detail is just lifted blacks, and once you reset them to black in post, a lot of the lower-contrast look disappears. In other words, I don't think filtering is going to really increase DR by that much, and certainly not if you do it in post since you are limited to whatever DR was recorded.

As for the notion that shooting without filters (other than ND's) and just working with the raw files somehow gives you the most flexibility, this is not completely accurate. If, for example, you wanted a super deep blue tone to the image, a blue filtered shot would color-correct easier and with less noise than a neutral shot because you aren't pushing the blue channel as heavily to achieve that look. So all not filtering does, in terms of flexibility, is give you the flexibility of not making up your mind as to what you want, but it does not necessarily mean that you will always get the best end result.

Now perhaps there are post software diffusion tools that do not impact the sharpness and gamma of the highlights, maybe the Tiffen DFX filters(?), I just haven't seen any at the post houses I've used.

 

[Mike Nagel]

David, thank you for your detailed response !

keep in mind that designs like Classic Softs and Hollywood Black Magics have a pattern designed to soften detail as opposed to just mist particles like a ProMist or Frost filter might have.

Can you elaborate on this ? do these patterns (in CS and HBM) introduce a slightly more uniform and therefore maybe more inorganic look ?

Now perhaps there are post software diffusion tools that do not impact the sharpness and gamma of the highlights, maybe the Tiffen DFX filters(?), I just haven't seen any at the post houses I've used.

Very interested in this and will run comparison tests, once my low con filter arrives !

 

[Björn Benckert]

Bjorn,

As usual, you are correct, but I'd like to point out a couple of things regarding the underwater part. I, too, have pointed out many a time that when shooting underwater, it is imperative to use optical filters to balance the light as much as possible before it hits the sensor, since that unbalance cannot be recovered later just cause the image "is RAW". However, an 085 filter, or any other warming filter, is not the best option for this task. There are filters made specifically for underwater work that have formulations much more suited for the actual conditions there, i.e.: with a spectral response that filters out much more blue/green than a warming filter. There are formulations for blue water and green water, where the filter repels either more blue or green, something an 85 does not do. I strongly recommend you use a purpose UW filter over an 85. Best choices are:

UR pro: http://www.urprofilters.com/ Crappy website, great filters, made of optical glass, mounted at the front of your lens, most thread sizes available. John Ellerbrock from housing manufacturer Gates can hook you up www.Gateshousings.com

Magic filter: http://www.magic-filters.com/ Again, crappy website but good filters. Gel strips can fit right over the OLPF. Found no many stores across Europe and US.

Trust me, either version will do MUCH better than an 085, but you can, of course, use whatever you like

Hmm, thanks for pointing this out. I mixed things up, and again I do not know much about filters. When I read your post I went back and checked our inventory list for the trip and realized that, just as you advice we ditched the CTO 85 (that would bring us to 3500K) and instead we have ordered something we (directly translated call) "Red/Orange UW filter" which is supposed to give us even better balance than the CTO 85... is that a correct choice according to your recomendation?

As it is, we shoot on a reef but also chroma key elements in a quite shallow pool and my concern is that I want it as neutral as possible before I do the composites and grade it all.

 

[Michael Notar]

Not ten minutes ago i just finished testing, BPM 1/4 and 1/2, Ultra Con 5 and PM 1. Shooting both in the shade with sun behind me, into sun, then allowing the filter to flare. my conclusions

BPM takes the edge of the contrast of the scene, the micro contrast and over all global contrast. These strengths are barely/slightly noticeable. You can compare and see a difference but not effect to be obvious to a view who doesnt know any better. They do exceptional things for portraits, interviews, medium close ups, in controlled lighting.

The ultra con reminds me of Cinestyle picture style on a SLR. It raises the blacks and lower tones a lot, but not much detail gain. It does reduce exposure of highlights and ups shadows obviously. I find once i have increased exposure, either in camera or post about 1/2 or 2/3 stop and grade it, i like it. Definately helps the DR of the image. Tts not magic, but it does help when in a pinch, esp backlight against a white hazy sky.

Side note- with a SLR or similiar camera, ultra con did not help make the footage much better. Too lossy of a codec for the heavy grading required.

Promist has the 80s portrait glamor look. Strong blooming of highlights, as the filter is a clear frost/mist. I got the #1 which is very strong. I would probably reccomend starting out with a 1/4 or 1/2. i am researching a etherial high key look for a dream sequence for a music video. I would also consider a BPM 1 or one of those lighter grades PM for the look.

As for color, I am definitely favoring using 81, 82, 80 series for color correction and leaving the sensor at 5000k. I even starting to use it with SLRs for photography. I have noticed less color noise even there and color tones are easier to correct. I like using the 81A with the low con or mist filters as its a free warm version of the filters for a vintage look. Ultra Con with 81A is really neat when the lens is flared.

 

[PatrickFaith]

When I ask this question to A level productions I don't get a either/or answers ... but it always turns out more complex then I was expecting.

Here are the filtering stages I have noticed on every A level movie I have asked detail questions on(sometimes it's not what they say it's way they do):
1) pre-lens filter (could be as simple as a UV or lens protection layer, but things like ND filters soften and reduce contrast by a degree even when a lowcon filter is said not to be used)
2) in-lens filter (I think most cinematographers if you call them on it, actually prefer filters that are embedded in the lens and give a specific look)
3) post lens filter (i.e. like the arri filter, nets, new RED nd filter, etc ...)
4) sensor filter (on the actual sensor)
5) sensor storage filter (i.e. wavelet, etc ... performs a filter function)
6) monitor filter (this isn't in the actual path to storage, but since people will change the shot based on the monitor filters, it does effect things)
7) DI filter (normally to get things to something like a 10 bit DPX, these do some perty massive operations )
7) compositing filters (every A level film where I have asked detail questions has serious compositing filters, normally to integrate the vfx)
8) grading filters
9) distribution filters (these do some heavy damage, it's one of those things were the best guys at this are the least appreciated)
10) projection/display filters (these can be good, but are often disastrous)

I btw did a little study of the cost of in-lens filters (i.e. like a cooke s5) vs doing that filtering in post (i.e. shoot with a still lens and then simulate a s5 in post). The cost numbers came back horrible(i was expecting the still lens cost to be acceptable - i was completely wrong), in that it was with todays computer time about 1 hour of processing per frame, and about 2 hours of human time per second shot, to simulate a cooke s5 with something like a chain of nodes in nukex with a cp2(nukex has the functions to simulate the s5, but you have to create full lens solves, and mix those solves with full pixel depth maps, and a lot of that requires per frame tweaking).

So with all these ten steps, there are significant financial implications on what filter is applied when and where, I get a feel from cinematographers it's like running a bit of a gauntlet - and they want to do some filter up front to assure the vision of the film is consistent.

 

[David Mullen ASC]

If someone does some halation comparison tests between doing it in-camera and with post software, I'd be interested if the waveform showed any increase in peak brightness with the post technique.

 

[Dr. Sassi]

I think the Epic Monochrome might improve image quality in certain situation with filters. Once the signal is monochrome, the differentiation is not simple anymore -- if at all.

For the RGB the situation is more complex. I haven't found a generalization for it (I work on that theme since a while -- for my book). With all imperfections of a generalization: if the camera can capture the full spectrum and dynamic, I would avoid filtration. Except when parts of the incoming light can't be differentiated later, or the look and feel is really defined. Polarization is certainly a practical filter, not a software option, and variable ND filter work based on that principle. A reason why I avoid variable NDs, it might have side effects. ;o).

Thanks for starting this theme. )

 

[PatrickFaith]

I think the Epic Monochrome might improve image quality in certain situation with filters. Once the signal is monochrome, the differentiation is not simple anymore -- if at all.

For the RGB the situation is more complex. I haven't found a generalization for it (I work on that theme since a while -- for my book). With all imperfections of a generalization: if the camera can capture the full spectrum and dynamic, I would avoid filtration. Except when parts of the incoming light can't be differentiated later, or the look and feel is really defined. Polarization is certainly a practical filter, not a software option, and variable ND filter work based on that principle. A reason why I avoid variable NDs, it might have side effects. ;o).

Thanks for starting this theme. )

my test was with monochrome.

If someone does some halation comparison tests between doing it in-camera and with post software, I'd be interested if the waveform showed any increase in peak brightness with the post technique.

It did require a reduction in the peak. Here is the nuke node (laplacian with gaussian, followed by log compression on the top), this also has to be done on the linear r3d(i bw got this from a major film but this exact node is mine so i can share):

set cut_paste_input [stack 0]
version 7.0 v8
push $cut_paste_input
Soften {
channels rgb
amount 1
size 21
mix 0.25
name Soften1
selected true
xpos -165
ypos -185
}
SoftClip {
conversion "logarithmic compress"
softclip_min 0.505
name SoftClip2
selected true
xpos -60
ypos -162
}

 

[Dr. Sassi]

my test was with monochrome.



...

Hey Patrick,

It must be a pleasure to work with several Epics (Color and Monochrome).

Thanks for sharing your information! Very much appreciated.

Best


(off-topic: I hope your movie is not eating all the time from your painting work away ;o) -- I sold my last hundred paintings in Europe in the past year and donated the money, I like to focus on digital purely since a while )

 

[Ketch Rossi]

Every one here knows my fillings about "IN CAMERA BAKED IN FILTRATION" but let me say it again, for me there is absolutely no way I would "PERMANENTLY" bake my images, and the only Filter I will always use will be the Polarizer, as the it is the only one that can not truly be replicated in post, as for all others it can, yet some filters are not easily replicated and some can't really be replicated at all, and in those very rare situation if it is for a particular need were I am to be 100%+ certain that that is what I absolutely need, and that there ail be no need to go back to the original RAW filked, then I would use it, other wise I would not!


The beauty of the RED EPIC cameras and the incredible POST flexibility and strength of the R3D files for me are invaluable, and I will almost NREVER alter the RAW in camera, but always in post, exception of the Polarization, and in absolute rare situation with a particular need with precise need for a particular look, I know I can not get in Post.


As far as Low-Con filters, I despite this type of filters and think of them to be absolute worthless and damaging to the RAW file, that if I was to see any one working for me try use one on any of my shoots, I would literally kick him in the nuts...