The Gentleman's Underwater Bubble Blower Thread

[Rudi Herbert]

Tom,

Sorry, but that only makes sense with non-raw, RGB cameras. True raw does not apply any conversions to the image and instead records light input values, so that all decisions including WB can be made later on after recording. Let's not forget that many of us who also do photography have been using raw way before RED came along and that indeed is the case with Nikon and Canon still cameras. Some formats like Canon raw, used on their cinema cameras like the C100, 300 and 500, bake in some values like WB and ISO to make the files smaller/more manageable. If Redcode does something like that, which has been stated not to be the case, then it would make a LOT of sense to set up WB during shooting and not later. Alas, besides RED's statements, every indication when working with Redcode points to the fact that it is, indeed, a pure raw format, even if compressed.

But irrespective of all that, the reason why setting WB underwater is only helpful to a degree, aka, a fool's errand, is because no sensor or color science can resolve the values needed to bring the channels into balance. The camera may correctly deduce that a temperature of X and a tint of Y are needed to balance the RGB (and remember, WB is nothing more than balancing the channels) but given the severe predominance of blue, and to a lesser extent green, coupled with the lack of red, it is simply impossible to achieve those values without incurring image degradation beyond repair. To that extent, I have concluded that the MX sensor cannot really go beyond temperatures of 12 000 K and tints less than - 35 without breaking the image apart, and in reality, the proper range is more like 8000-10000 K/-15-25. Which means, every time your camera asks for values beyond those levels, you would do well not to acquiesce and instead rely on other grading techniques to bring your colors back to life. Of course this applies only to underwater, on land, the MX and most other sensors can definitely deliver within those ranges, but that is only because the light is far more balanced.

Which is why, irrespective of whether raw or RGD, RED or Sony, etc, etc, I never bother with WB values beyond 8000 K, but that of course, is just me.

 

[Tom Hamilton]

As always Rudi, I defer to your expertise in these matters, as well as that of some very respectable professionals who contribute to this thread. Re stills and DSLR and solid accomplishment in the field, I have asked David Doubilet to chime in here on this thread, but NatGeo assignments being what they are, he is a very busy man. I admire his work greatly, and strive to emulate in video what he does in still, always another mountain to climb.

In theory, RAW should enable the shooter to frame, focus, iris for max DR, worry about the rest in post. There are, however, comments posted by some solid guys here that state, proper WB in camera gives you more latitude in post. Still trying to wrap my head around that one.

 

[Rudi Herbert]

The only way to wrap your head around it is, if it is true raw, then it makes no difference whatsoever, and if indeed there is a difference, then Redcode is not completely raw. It is simply one of those things that is either black or white, no shades of grey in between.

 

[Jeff Whitehurst]

Gentlemen, The following is a series of crops taken from 2 separate R3Ds showing just the Red Color Target on my UW color chart. Both files made at 70ft, seconds apart. For additional analysis, I isolated just the Red Channel for each sample.

The images reflect various preliminary options available in the post process using a file with no WB set in the scene, the 3rd image shows an untouched image with WB set in production.

Now, I simply ask you to judge, which has more color out of the gate, better luminance and less noise? Which approach would produce consistent, accurate color throughout a sequence of shots? Which file would you rather hand off to a colorist?

I would also point out that if either of these files were shot without a true reference for white balance, then options 2 and 3 would be impossible to achieve without guessing.

Hope it helps, Jeff

 

[Björn Benckert]

Björn, Thanks for your input, but with all due respect, (and I have much of that for you) telling me "you are wrong" doesn't explain why my chart indicates that I'm right. In the first image I posted, I did exactly what you suggest, I adjusted the WB in post to an accurate target. Still the channels, especially those that are deficient at depth, are weaker, so setting WB in post doesn't bring back data that was lost.

Further, the statement made earlier that WB affects nothing prior to deBayer directly contradicts the information provided from RED.

So, in common practice, RED UW shooters are shooting with WB set to incorrect values, WB is set in post to inaccurate targets (or not set at all), and as I demonstrated they do so with deficient data. Still, one after another comes on REDUser and complains about how "crappy" RED footage is UW, and time and again I and others demonstrate that correct WB is essential in the post processing color workflow.

I might be completly wrong but all my reading, testing and shootings indicate that wb in camera only adjust the metadata tint and kelvin slider, nothing else, exact same parameters that is available in rcxp and that can be altered by using the color picker....

if there is any other parameters that change then thats a huge suprice to me as i usually shoot in layers and never care at all if i change wb in camera or even swap looks file while shooting the different passes it has allways been possible to apply a comon seting in post and the two differently shot layers have had a perfect match...

So basically what you say goes right against all of that...

to me its simple expose so that you have all rgb channels filled up and make sure that what is important to you lands at 70% (read skintones) then you got most out of the sensor. What a lot of people do not understand is that if one or more chanels are over or under expossed then your image loose quality. Most underwater shootings leavs you with a starving r channel so then you need to filter to hold back b and g. and ofcourse open up to bring all three up to a decent level. wb can only slide your values ariund a starved channel can be brought up but then as those photo sites only got hit by a low number of fotons they got less consistency so what you get is noise on red you do not only get noise you also stress the compression with this noise, and that degrades your image even more.

Also I do not completely understand what your images above actually shows I understand that they somehow shows difference at K12000 when that value is set in camera or in post, is that what you are saying that they are?


Another note, To me it's much better to use a looks folder for UW shooting preview / default seting than just adjusting the Kelvin and tint. Make a looks setup in red log and S shape your luminans curve and bring up your red curve at the bottom dark end and even introduce a little green down there. Then you punch the red signal where it's needed and not across the board as the magenta is not really lacking in the highlights only low lights / deeper depths so then you get a better look.

Thats whats done here sadly I clipped those cuves in my screen dump for the image to the right, but that correction is very possible to use in a looks folder on the camera.
http://classic.syndicate.se/Files/~usr/hawk/RAW_MAGIC.png

And as you see on the histogram this images was still about 20% low in the r channel. So more orange filter would be better as blue and green are pretty much at the same level at 5600K

 

[jim kozmik]

TO JUST BUMP in here..most of this stuff is centred around the red epic x camera underwater I AM only assuming..QUESTION is this same problem occur with my old RED ONE MX camera underwater....please advise me..FRASIER SAYS HE USED TO USE RED FILTER in shallow florida and bahamas waters..kozzzz

 

[Jeff Whitehurst]

Also I do not completely understand what your images above actually shows I understand that they somehow shows difference at K12000 when that value is set in camera or in post, is that what you are saying that they are?

Hi Björn,

I removed the images showing WB set to 12K in post, I guess that just confused things. The first set shows what happens to the red target on a chart with WB set to default, the second shows WB set with the eyedropper and the third is WB set in the scene, that's it.

The way the post reads now is all I really wanted to show, 3 ways to post an UW clip using WB to correct color casts as a starting point. As I demonstrated earlier, sampling the red target in RCX reads much higher when WB is set in the scene. Why this is, I don't know for sure, all I can say is I performed this test three times, it always resulted the same. I was surprised as well that the original data was affected, but it clearly is stronger. I should stress again, I changed nothing other than the way that WB was set, everything else is set to default values.

It has been argued on many occasions on this forum that setting WB in production doesn't matter, because it can be set later. Others have gone further to say that WB doesn't matter at all, but as I pointed out earlier RED says it effects everything down the chain. The worst problem is, most of these UW shooters have no way to even set the WB other than a guess or using the eyedropper on the sand or whatever. More than a few have come on here bashing RED cameras and complaining how bad the color is, claiming heir images are crap, and in the end discover that their data is there they just don't know how to get to it. In the very least, my test provides a reliable, non-destructive and free method for improving the files off the camera with minimum worry. That's the main point I have here, if you want decent looking images straight off the camera or are recording externally with a baked in look, set the WB accurately and you have a decent base look.

Regarding your image, (which is very nice BTW) I think you made a good call using a filter as your histo looks good. In addition, since you have strong red elements in frame there's a nice signal there to work with. I've noticed that many problem images shown here have some common issues. The red channel is in the dirt and boosting the red channel breaks down the shadows or there's nothing with any red color in the frame, so they have a hard time making anything pop. Your advice about manipulating specific aspects of the red signal is sound, as I'm seen that many complaining about noise in the red try simply boost the entire channel, which is going to give a magenta highlights every time, especially if there's no red signal and no WB set.

That's about all I've got today, I'm off to try some egg nog, Merry Christmas everybody

 

[Johnny Friday]

Well, while I don’t understand why I’ve been dragged into a contest of skills there was no need for, and while I have managed to survive every time it turns out somebody is better than me at something , this statement nevertheless puts a wrong spin on things. Given that Paul has continuously repeated he cannot see what the problem is, and I have done nothing but insist the issue is a serious one, it may then be inferred that this is nothing more than inferior color grading skills at play and nothing more.

Skills have nothing to do with this problem.

Rudi: in plain ingles....I was merely paying you a compliment....Hold your latin temper back....I was actually suggesting that you know how much goes into coloring an image. And Paul while not seeing the problem with the magenta....he still knows that art of color science better than most posting in this forum, if not more than all of us. I can only say from personal experience in working with Paul for about ten days doing nothing but color work...to say he knows far more than basics to bring an image from just good to exciting using a variety of tools i never would have used nor do i care to use.....You are one of those as well that knows far beyond any basics. In short, i'm merely vouching for Paul as someone that can look at an image most of us would look at and call dull and for some reason he sees the possibilities---based on what he can do with it....NOT at all implying a test of skills between you and Paul. Paul is an interesting individual to say the least.....i would throw things to the side while he would work them later and show me a different image.

Anyway....no test of skills...only Paul truly sees things differently and he's seen quite a bit of underwater footage from me and made me consider more what can be done under the hood. Ok, that said, yes we still face massive issues with Magenta in images and like any image, if there is not enough light, it's going to fall apart. The blue background seems always to be tricky to work with in RED footage (from my experience) and to get it really right seems to take secondary work--mostly work i don't want to do....I'd say it's good enough. BUT, not perfect..AND it can be made to look great given the ability to do that secondary work....much of that i'm not willing to do to a point...but one hour of footage and 1.5 hours post is my limit.

....anyway, some of the points you made about the image i agree with and one about redcode we just have no control over. And i agree with you about the image being rather lackluster...but it can look good after a pass by a seasoned colorist as we can see. And i think all of these passes were done in RCX which in my opinion is just not good enough to do that work anyway. That's where i take some divergence from Paul's last statement. I think it gets you in the park yes...but not to home plate.

My apology if i made a statement that made you take offense. Was not at all what i had intended.

 

[Lauri Kettunen]

Björn, Jeff is saying the in-camera WB matters, *if* the color temperature is far from 5600K. An explanation could be, there has to be enough data for debayering on each channel. If one is far away from 5600K perhaps the camera does some cross-talk between the channels to capture a proper signal on every channel, and this then is counterbalanced automatically in post. Whatever is the explanation, evidently there are conditions in which a bayer sensor requires special techniques to capture enough data.

 

[Johnny Friday]

Jeff: to be certain what you are saying regarding your reference chart and a few questions:

1. Are you stating you are shooting the reference chart for post work later OR are you saying you actually set WB in camera on that chart? In your first image of the chart you have some a backfocus or focus chart with some rectangles of differeng color values...My question is if you set WB to that---how do you know WHERE you are setting WB ? if you are not using a fully 18% gray chart to set WB underwater and in Scene--how do you know WHERE the WB sample is being taken?

2. Or...are you just shooting your u/w chart to later sample (use the picker) to choose an area to set WB?

....a reference is always great and like a lot of others we may all use some tricks....i just use gray and white tape on my fins to shoot them before or after take....

I'm genuinely curious as to how you can achieve a fatter red channel by setting WB underwater and not able to get same results later...I've done both and i my tests i found i could just do it later in post....and use a reference. As Rudi points out and as you likely know as well, with enough time shooting and working behind a computer, it becomes second nature to know what values a certain RED image will require in your basic settings. Where i find it becomes challenging is shooting up to light or down to sea floor. But in most cases if there is substantial light and regardless of night or day, you can get pretty darn close by eye and scopes. And i feel like anyone can do it given they put in enough years of working their images---but therein lies the big jump.....takes time and effort. There are a handful of seasoned shooters on this forum that i think get that easily and i'd bet you are one of them.

That said, i'd be very curious to see how that red channel can become fatter by an in water WB. That in now way means i ignore it...i'll look at water while in water and set WB to my liking based on the days water conditions--something i prefer to do and then balance it all out later in post. In the end, it's obvious we face a big magenta issue and this is not just an underwater issue with RED. I'll see if i can dig up an online article about it....but it's there. I believe RED has taken notice and this new sensor seems to account for a much cleaner image.

 

[Rudi Herbert]

Johnny,

Didn't know I had much of my Latin temper left after having been slowly, albeit lovingly, neutered by my European wife for the past 14 years...Anyway, I will hold back and, worry not, I like you enough to both not be bothered by your words even if they were meant as a comparison and to want to stay on your good side.

Jeff does bring some interesting points here. First, I never said WB does not matter, simply that with raw you can worry about later. I also said that I always set WB in camera to about 7000-9000K while I'm shooting, and leave it untouched afterwards because, in my experience, anything much higher than that degrades the image at a global level anyway. But of course it would be foolhardy to, say, set WB to 2500K and never touch it during post. All I can say is that people who do that really have a very poor understanding of image processing and probably should not be shooting with a professional raw camera like RED.

Now, I asked Jeff to send me a few of those sample R3Ds for me to look over, and in fact, it would be good if people like Bjorn and Paul got a hold of them too. I am not doubting Jeff's word, but something like what he claims can fundamentally affect the notions of Redcode we've been working under for the past 6 years and should be corroborated by other people. I hope Jeff will indeed share those images. But even if somehow Redcode needs WB to be set during acquisition to deliver better images, that still does not answer a lot of other questions:

- I have shot with my customary 8000K and -15 tint in shallow water on occasions where, coincidentally, another camera chose values of 7900K and -18 tint while WBing on a card, and still the images were terrible, exhibiting the same issues they do in many other conditions. I was shooting basically at the "right" WB and got no benefit from it whatsoever.
- Again, if we were to WB at depth, and come up with values in excess of 12000K and -35, something the RED does quite frequently, then irrespective of when the WB was set, you'll end up with an image that is severely compromised because, as I explained before, no sensor can really resolve those values without triggering impossible gain levels in the red channel, the result once again, a terrible image.
- So, if this WB set up at depth works, it still can only work within certain levels, and then all WB readins at depth are nothing more than an approximation anyway, for it you position your white or grey card 1 ft in front of the camera, you'll get vastly different results than if you put it 5 ft away, and so and so forth. Going with a pre-set value like I do may not in fact be much different.

Those of you who have not looked at the R3Ds posted by Darek, I encourage you to take a look at them, those images are incredibly awful, very similar to the ones posted by Greg F. from Hawaii. I know Paul took a pass at them and posted some small JPEGS, but I would ask that Paul spent some more time and try to get those images to look their absolute best and let us take a look at a full size TIFF or PSD of it. If he really manages to endow those images with accurate, saturated colors and no or very little chroma noise, then I think we could present his results to Graeme for him to figure out how to translate those manipulations into color science tweaks. Those are are the type of images I would really throw away after just balancing them on RCX.

 

[Tom Hamilton]

Fascinating discussion here, the thought has occurred to me, as I'm certain it has to everyone else, we may just have stumbled upon the RedCode flaw that has been responsible for the magenta ghost haunting Epic.

 

[Dave Blackham]

I hope Red are away enjoying the festivities right now as we all should. But…

Can some one please describe how Epic is calculating WB. i.e. is it just degrees K or is it looking for balance between RGB or what.

Im fairly sure the numbers it comes up with underwater, as I think Rudi is saying, (please correct me if Im wrong), are completely incorrect due to the change in light spectrum at depth.

So is this likely to be leading to 1. A change in Colour science. 2. A filtration issue, 3. Something with in sensor design that may not be able to be cured, To fix the issue.

Thanks to everyone discussing this and looking for a solution.

 

[Johnny Friday]

Dave, Rudi, Jeff, Tom, et.al.
I would be fairly confident that RAW is RAW and be surprised to find that there is something baked to the footage. There have been years of testing by enough skilled DP's shooting some big features and doing some very extravagant camera testing.
...As was pointed out by someone earlier....Rudi perhaps--many cameras exhibit this Magenta cast at some point when being ND'd down.....In fact this Magenta Phenomena is not just evident in underwater footage, but well known to topside shooting when applying ND's to lens and seeing how far open you can shoot before a Magenta cast appears and it just too much to handle. I've seen some comparison tests with RED and other cameras.....the very latest test i saw was by the DP Anthony Dod who shot Rush....he did some very extensive testing with EPIC MX and Arri Alexa and i believe a few other cameras (sadly i cannot find that article/video) and he points out where the magenta cast appears in footage when shooting open and ND'd or at night. So this is well documented and known. We of course see it to a far greater degree due to the fact that in some cases (my thoughts only or guess) the water is acting as a filter or ND to some extent and we are seeing exactly what these topside tests reveal with an ND filter some water acts as more a filter than others......based on ??? a prevailing thought is UV....Regardless, EPIC MX shows more of this magenta cast than compared cameras such as Alexa and ??? Anyway, i'd be willing to bet that RED is well aware of it and in Dragon has worked to reduce this Magenta affect. We in underwater will of course deal with it more often. Is this an issue only representative or RED sensors? likely not since we see it across the board. Perhaps an issue with CMOS and then to what degree or soup does each camera company stir up their recipe to create a better color science is likely what's happening now.

 

[Darek Sepiolo]

I second Rudi and encourage you to manipulate my R3D file. It was shot on Palau, c 10 meters deep, sunny day, clear water. I white balanced the camera on sandy bottom probably being half a meter from the sand. This was the same procedure I used to do for 3 yrs shooting for Discovery with my old EX1 and having no problems with colors.

http://dive-away.pl/dopobrania/A003_C013_05150R.0000387F.R3D

 

[Johnny Friday]

Derek, just as a sidenote......and i'm not picking on you, but some shots just do not lend themselves as a keeper....for whatever reason. Composition, distance to subject etc....I have plenty of those and they just go in the dumpster regardless of what it was shot on. But also, you are doing WB on the white sand? This does not lend to a proper WB anyway and i would suggest either doing as Jeff does and WB on a chart or set your target in camera and deal with it later. I agree with Rudi on this one in that setting WB underwater to something other than a chart leads to numbers that can be so far out of whack that it's not worth doing unless you have a chart.

post note: in fact after looking at your settings in camera it's obvious that the image should look terrible with a WB of 24,477 and tint of -38.8.....i made a pass at that image and can't get much better either.....but there are so many variables as well....

 

[Marc Thorpe]

Derek, looks like Ulong Channel. A great dive, IMHO the best in Palau.

 

[Darek Sepiolo]

Derek, just as a sidenote......and i'm not picking on you, but some shots just do not lend themselves as a keeper....for whatever reason. Composition, distance to subject etc....I have plenty of those and they just go in the dumpster regardless of what it was shot on. But also, you are doing WB on the white sand? This does not lend to a proper WB anyway and i would suggest either doing as Jeff does and WB on a chart or set your target in camera and deal with it later. I agree with Rudi on this one in that setting WB underwater to something other than a chart leads to numbers that can be so far out of whack that it's not worth doing unless you have a chart.

post note: in fact after looking at your settings in camera it's obvious that the image should look terrible with a WB of 24,477 and tint of -38.8

No problem Johnny I know it's a bad shot and I'd never use one. I kept it only as a record of bad colours.
Palau was my first trip with RED and I just tried to repeat the same procedures that worked for me for yrs. I did for Discovery 7 episodes each was 45 minutes with EX1 and there was no single complaint on WB I used. And most of that was done on sandy bottoms. I used to have a chart but after few tests I gave up this method. So the WB of 24,477 and tint of -38.8 are just the results of me doing the same thing that worked with my previous setting. I was quite supprised with the bad colour renderings but at that time I didn't know about the magenta issue. I had different limitations with EX1 and my job is to find the solutions. For RED it was not to shoot that kind of scenery. I could get so many other great shots that I neglected the problem until I found your discussion here.
I'm not a colour scientist I just point the camera and try to get the best shot possible. I've already posted earlier the below picture. Very similar UW conditions same WB procedure just two different cameras. I expected similar results but couldn't get it. So as I mentioned I gave up those kind of shots. But it'd be great if RED finds a solution for the problem.

Marc, I don't think it was Ulong. I can't remember the spot. But you're right Ulong is great dive

 

[Tom Hamilton]

[QUOTI hope Red are away enjoying the festivities right now as we all should. But…

[/QUOTE]

hahaha Dave, someone needs to get a feckin life here. :smiley:

That having been said, we may just be on to something here, Xmas won't derail this train...

 

[Marc Thorpe]

No problem Johnny I did for Discovery 7 episodes each was 45 minutes with EX1 and there was no single complaint on WB I used. And most of that was done on sandy bottoms.

Cheers Darek, I told the wife I was trying a new WB technique, I think the bruising will disappear in about a fortnight...