Expose to the right

[DiaAzul]

The most important element here is experience in terms of judging the dynamic range of a particular scene and compensating accordingly. If the scene is uniform in brightness then exposing to the right is possibly the way to go. However, where there are particular hotspots in an image then space must be left to the right of the histogram in order to accommodate the brighter areas in the image. Good examples where this margin is required is Didger's specular highlights when shooting into the sun, but also in nighttime photography where there are bright lights exposing to far to the right can destroy the quality of the image.

Another factor to bear in mind is, as far as I can work out, the histogram is created based upon a subsample of pixels in the image. Therefore, it maybe that you are not capturing all of the hot pixels in the image. From my point of view the histogram is an approximation of the distribution of luminance values in the scene and should act as an additional tool to the 'photographers eye', the exposure meter and any other tools available for determining correct exposure. No one tool has ever been proven infallible and a good photographer is one who knows the limitations of the tools and how to compensate appropriately for a particular scene.

Though not available on the 300D, I find that switching to spot metering for high contrast scenes enables a more accurate determination of exposure to be made (one area the the 1-series cameras have an advantage).

If photography didn't require experience to cope with difficult situations then there would be no need for professionals.

[Jonathan Wienke]

You can test your camera to find out the exposure interval between the histogram's clipping indication and actual RAW clipping.

Digital Exposure And Metering Strategies

[61Dynamic]

Dinarius,

There are three types of historgrams. RGB, individual channel histograms, and Luminosity.

An RGB histogram (like what you see in RGB mode in Levels or ACR) shows the position of every pixel from all channels in the scale from black (0) to white (255). This histogram will show clipping of any sort, but it won't indicate what channel is clipping unless it's in ACR or the histogram pallet when set to show channel colors.

The individual channels as you might expect are the same as the RGB histo except for only a single channel.

The Luminosity histogram is the thype of histo camera shows. It is essintially an average of the three channels showing an images luminosity (whith a bias towards green). If a single channel is clipping, a Luminosity histogram will make no indication of it (unless that channel is green - usually). Generally, if there is two or more channels clipping, it will be indicated in the histo.

[Dinarius]

On the related issue of density........

I use Kodak Color Seperation guides in all my artwork shots.

Is there any way I can select the middle grey in those images and set it to 128, thereby making the entire image the correct density?

That would be very useful.

I use PS Elements only.

Thanks.

D.

[Leigh]

Though I generally shoot RAW with my 20D set to Parameter-2 ( no in-camera processing), I have a custom Parameter Set in which only the Contrast setting is backed fully off, which helps to prevent the highlights from being blown out in very contrasty scenes.

   Leigh

[thierrylegros396]

Using ETTR technique is not enough, especialy with the "not-very-accurate-Histogram" of the EOS 300D !

In high-contrast scenes, you can have a lot of burned pixels, although the histogram shows a safety margin.
Light metering seems to be part of problem with EOS 300D.

Anyone who have a lot of skills to improve the method ?

Have a nice day !

Thierry Legros

[didger]

Yeah, I'm no longer such a great shoot as far to the right as possible fan.  You can not only blow out individual colors sometimes without even a hint of this on the histogram OR the flashing warning blown pixels, but you can also sometimes get not only blown pixels but total unsalvageable color noise garbage areas under certain circumstances.  I've had two series of shots pretty much totally and irretrievably ruined where I had to shoot into the sun and there were very bright specular highlights on water and/or ice.  There was no hint of flashing pixel warning or from the histograms.  If I had slightly underexposed I probably would have had no problem, or at least not an unsalvageable level of this problem.

You can deal with a little shadow noise if you have to, but badly blown highlights or this color noise garbage are harder or totally impossible to deal with, so I'll go for center histogram from now on for most cases and histogram to the left if I have any doubts about possible problems at the right end.

[Jack Flesher]

If photography didn't require experience to cope with difficult situations then there would be no need for professionals.

Well said!  I am going to steal this from you and use it regularly

[61Dynamic]

It's just a matter of getting a knack for it. With my 300D (and the 10D and 20D I also use) I find that it actually doesn't blow out as much as indicated in the histo (moreso with the 10D). There are exceptions to this of course; esp. if you are shooting in broad daylight. Single or dual-channel blowouts are always a gamble with any camera that only shows a luminosity histo. For this, you need to pay attention to how saturated the subject is. Reds always blow out sooner than other colors.

It's a matter of knowing how the camera will react based off the light and the subject. With practice, you should be able to see a scene and know what range of light the camera will be able to capture and what colors might clip before taking a frame.

[didger]

Shooting digital is so cheap.  I figure once the tripod is set up, if it looks like a sure keeper in the viewfinder and there's any doubts about anything, shoot it every which way, maybe twice.  Better that than gnashing your teeth later.

[Paulo Bizarro]

Experience is all. The histogram is but another tool that we have available. In my first outings with the Powershot digicam, I have included post-review histogram checking in my workflow. Sometimes, I find that the Blue channel is overexposed (the PS Pro 1 does not have individual RGB channels).

I just do the thing that I have done for 10 years: trust mt experience in judging the rendering of the scene, and bracket. Even with slide film, which is actually not that expensive  

[didger]

didger doesn't say whether he is shooting RAW or JPEG; but, if it is RAW and you are getting blow outs despite the histogram not indicating this, then this totally contradicts both my experience and the central tenet of the article

I only shoot raw.  I haven't ever had the problem other people have reported of particular colors blowing out with no hint of this in the histogram, but I don't doubt that this can happen.  My problem was definitely something quite different from a normal "blow out".  Blown highlights just give you 100% white pixels with no detail and you can set the camera to show you flashing blown highlights.  What I had was specular highlights so bright that some sensor elements failed altogther and I got color noise and there was absolutely no hint of anything wrong on the histogram or flashing blown pixel warning.  I know exactly the circumstance when this happens and I think in the future I can avoid it completely by NOT exposing to the right or maybe with a polarizer.

If you mainly shoot artwork in a studio with controlled lighting you'll certainly never see this weird color noise phenomenon that is quite rare even in outdoor shooting.

[Dinarius]

Daniel,

If what you say is correct, then didger's experiences are borne out  and it begs the question, what is the point of using this EttR technique at all? What you and didger are saying seems to be undermining EttR in a very fundamental way.

Take my example of shooting artwork for reproduction: it is conceivable that one channel may be burnt out without it showing either on the histogram or flashing on the screen, just as didger wrote. What can one do? Bracket, I suppose and check the channels later in PS. But, that's a bore. It adds time to the workflow and in my line of work, where one might have 30-40 shots to correct and process, that's a lot of extra time.

Are there cameras that allow channels to be checked seperately at the time of shooting?

D.

[61Dynamic]

Yes, the luminosity histo on cameras is the baine of ETTR. If you are not carefull, you could in fact clip info an not even know it.

This, as digler pointed out, is where experience comes to play. The MkII cameras do have individual channel histos (although some debate the accuracy of them saying info clips slightly sooner than indicated) but knowing how the camera will capture that info base off what you are shooting and the lighting is indespencable.

You basically have four options:

1) Continue what you are doing taking the risk that you might be cliping information in a single channel. If you are pleased with the results and find you aren't loosing any important data using ETTR under controlled light, then more power to you.

2) Don't use ETTR with certain saturated subjects. Use a "correct" exposure to get the best balance of information.

3) Under-expose some subjects 1/3-stop. For this one you need to weight the costs of such. Is the info you are preserving from getting clipped more valuable than the info you are loosing by under-exposing? If the vast majority of the subject contains a highly saturated color with needed tones/texture in it then perhaps that's the case. If the satuated color is minimal, then probably not.

4) Bracket. As you pointed out, that's a bit of a nucance but it may be the best choice for mission critical images.

No one can tell you which is best, only you know that. Do a little experementing and see what fits your needs.

Also related:
Restore Those Clipped Channels

[tmcope]

I know the last post on this topic was a little while ago, so sorry if I am bringing up an out-of-date topic.

I think there is a problem with this "expose right" argument (as in expose to the right, not expose correctly) that Michael puts forward:

http://www.luminous-landscape.com/tutorial...ose-right.shtml

I think the problem is this:

The F-stop system is a non-liner exponential system
CCDs record brightness in a linear fashion.

However the histogram is linear (isn't it?). It shows 256 (in 8 bit mode) or 65536 (in 16 bit mode) levels of brightness in a linear fashion. In other words, the histogram shows LEVELS not f-stops. This means that the right hand fifth of the histogram does not represent a full f-stop of dynamic range. The brightest f-stop of dynamic range is represented by the top half of the histogram. The next f-stop is represented by the quater of the histogram below the half way point, the next f-stop by the next 1/8 and so on.

So, if you don't expose the top 1/5th of the histogram you are only loosing as many levels as your histogram tells you you are loosing (i.e. 256/5 or 65536/5).

What you see on the histogram reflects the linearity of the sensor. So I think this argument for shooting right might be a bit overblown (pun intended!)

[Jonathan Wienke]

I think there is a problem with this "expose right" argument (as in expose to the right, not expose correctly) that Michael puts forward:

Read this article. It builds on Michael's article, and addresses your issue. As to Leigh's comment, the only relevance it has is that camera JPEG processing settings affect the creation of the histogram, but do not actually affect the RAW data. The best thing to do is to find camera settings where the histogram and RAW data match as closely as possible. The article describes how to do that.

[Graham Welland]

If someone's paying for your work then I would seriously consider moving to PS CS.  In your case the curves grey point selection and adjustment layers alone would solve your problem.

[Jack Flesher]

It's the double-edged sword of using ETTR...

In concept you want to expose as far to the right as possible without blowing any single channel.  A monochrome histogram may look fine, full tail visible in the LCD histogram, but you may in fact have blown the uper end of the red or blue channel and it doesn't show.

What we really need is the ability to magnify the LCD histogram -- an RGB histogram -- in the 235-255 range so we can see exactly where the channels end.  Maybe we'll get this in the 1DxMKIV cameras...

In the meantime, I suggest leaving visible daylight between the right edge/tail of your luminence histogram and the end of the scale.  With more complete 16-bit support available in CS, and better noise management from the DigicII chips, you can get away with a more centered histogram and still maintian good shadow detail.

My .02,
Jack

[Eric Myrvaagnes]

No one tool has ever been proven infallible and a good photographer is one who knows the limitations of the tools and how to compensate appropriately for a particular scene.

I can't resist passing on another Edward Weston anecdote I heard in a Minor White workshop. Edward and Ansel were out photographing together. At Ansel-the-Father-of-the-Zone-System's insistence, Edward had aquired an exposure meter (a Weston Master V, I believe -- no relation). Here's the way Edward used the meter: He would wave it around wildly, then look at the dial and mutter "It's wrong!", and then give the exposure he had planned to give all along.

Of course, Edward's eye was as close to an infallible exposure measuring device as we are likely to find.  

[Dinarius]

This is a very interesting thread.

My experience would be very much that of the article to which Jonathan has kindly provided the link and not that of didger.

i.e. Shooting RAW in Adobe RGB, I have sometimes found that even though I have exposed fully to the right, the result is comfortably short of any clipping.

didger wrote, ''You can not only blow out individual colors sometimes without even a hint of this on the histogram OR the flashing warning blown pixels''

I must say that this has never been my experience. didger doesn't say whether he is shooting RAW or JPEG; but, if it is RAW and you are getting blow outs despite the histogram not indicating this, then this totally contradicts both my experience and the central tenet of the article referred to above, namely, that in RAW with the histogram as far to the right as it will go without clipping, there is still room to play with; fully 2/3 of a stop according with the 1D, according to the author.

I may be stating the obvious, but one point which Michael clearly makes in his article "Expose (to the) Right" is that an image shot using this method will *look* overexposed on the monitor.

He writes, "Now of course when you look at the RAW file in your favourite RAW processing software, like Camera RAW, the image will likely appear to be too light. That's OK. Just use the available sliders to change the brightness level and contrast so that the data is spread out appropriately and the image looks "right"."

I shoot a lot of artwork for catalogue. My main issue is colour, needless to say. I always expose to the right and , occasionally, I have to be very careful that tiny specs of white paint in an otherwise dark painting don't blow out. These can be almost invisible both on the histogram - because they are rendered only as a tiny line on the right hand toe of the histogram - and as blinking pixels on the screen.

I always include Kodak colour strips in the shots and these always look washed out on the monitor. I find the colour from my 20D RAW files processed in DPP to be extremely good. Much better than film. I find that only brightness and contrast need to be adjusted. But, getting the right density on the Kodak grey areas is tricky. If anyone has a formulaic way of getting this right, I'd love to know. Currently, I am adjusting the grey until it looks right on my calibrated monitor. But, I'd like to be more scientific about it. Please!! ;-)

D.