exposing to the right (again)

[Phuong]

i remember there was an article/tutorial on how to know by how far you can expose to the right with your specific camera. basically, make a few raw captures of a same scene with increasing exposure, and then check their histogram in ACR, then you'll be able to determent the "right" way to get the "right" exposure with your camera.

however i fotgot the specific details. i've been trying to search for that topic (both on forum and on the main page, and on google) but have no succeed.

does anyone remember how to do it? or better yet help pointing me to that topic? i'd appreciate very much. thank you!!!

[Ray]

It's a very contentious issue because the histogram cannot distinguish between specral highlights and slightly more than specral highlights. It's all about getting a balance between desired shadow detail and desired highlight detail, or maximising the dynamic range of the camera. If you haven't shot far enough to the right, the penalty is increased noise in the shadows which you could have avoided by a bit more exposure. But sometimes a bit more exposure will blow detail in highlights which you might want to preserve.

There's no formula for getting it right, as far as I know. It's basically up to the photographer's judgement and experience. When in doubt, use AEB. That's what it's for.

[madmanchan]

Here's what I do.  I shoot in RAW mode but set the in-camera image parameters to a low contrast setting:  -2 on my 5D and -1 on my 1D Mark II.  This causes the histogram on the back of the display to match fairly closely the histogram that I see later in ACR, my RAW conversion program of choice.  In other words, if I see blinking highlights on the back of the camera's LCD, then I'll see an overexposure warning in ACR, and conversely.

Then I spot-meter the scene and look for the brightest region that I wish to keep detail in.  I place this bright region at about +2.5 stops over the meter reading.  I take a test exposure.  Most of the time, nothing is clipping, but the histogram is pushed to the right.  If it's not pushed all the way to the right, I increase the exposure slightly and try again.  If I accidentally clipped something, I back off.

Make sense?

Eric

[Dale_Cotton]

Phuong: I think the document you are looking for is this one:
Expose Right

This explains the theory more than the practice. My own practice is to meter the brightest portion of the scene then subtract 3 stops, which happens to be the latitude of my camera from what the spot meter sees as "middle grey" to the point just before highlights start to blow.

Specifically, I work in manual mode with the in-camera spot meter on then point the camera at the brightest portion of the scene (usually in the sky), watching the exposure read-out in the finder and thumbing the shutter speed dial until the read-out says +3.0. That's my exposure, so I can now grab focus then compose the shot. This routine is simple and fast enough that I use it for all non-action shots.

Lacking an in-camera spot meter, one could review the shot with the in-camera histogram and/or blinking highlight indicator; alternatively, one could spring for a (possibly second-hand) handheld spot meter. Another approach might be to use your camera's automatic bracketing mode in conjunction with whichever metering seems most reliable.

Note that +3.0 is what works for my camera; you need to find the equivalent number for your own camera.

Ray: I note with interest your objections to this approach but can only say that it works for me.

Eric: your response appeared as I was composing my own. Interesting overlap.

[Chris_T]

Eric and Dale, the manual spot metering technique is exactly what I use for my slides to preserve the highlight details (sometimes at the expense of losing the shadow details). I often wonder how this is any different from "exposing to the right" on a digital camera with manual spot metering. Such an exposure will push the histogram to the right without clipping. From there, using raw conversion can extend the dynamic range, which of course cannot be done with slides. It would be nice if the "expose to the right" tutorials can clarify this.

[Gary Ferguson]

First you need to establish the latitude that you have (some people claim this varies with the ISO setting but I've never found a difference) there's different ways of doing this but the simplest is to fit a long lens and aim at a grey card, uniform area of blue sky (with the sun behind you), or a plain coloured blank wall. Then increase the exposure by third or half stops until it blows out. Knock off maybe a third of a stop for safety and the difference between the original and final exposure is your over-exposure latitude. Usually it's somewhere between two and three stops.

You've then got three main ways of proceeding.

If you're handholding and shooting fast changing, dynamic scenes then you've little option but to trust your camera and use auto-exposure. If the scene includes very bright areas then you may dial in some compensation, or meter selectively from a part of the scene that's a bit brighter than the average.

If you're shooting in a slightly more considered way (think travel rather than documentary/street photography) then you can spot meter from the brightest part of the scene where you want to retain detail then open up by your previously calculated exposure latitude.

If you've got still more time to construct each shot, typically landscape with a tripod mounted camera, then you'll want to take test exposures and adjust the exposure until you've squeezed the histogram right up to the right. I find that it's best if your first "ranging" exposure is targeted to give you an underexposure rather than an overexposure. This is because, with experience, you can usually judge from an underexposed histogram exactly how much more exposure you need to take it all the way to the right. If your first test exposure is over exposed the histogram gives you no useful information beyond the fact that there are blown out highlights.

[Ray]

Here's what I do.  I shoot in RAW mode but set the in-camera image parameters to a low contrast setting:  -2 on my 5D and -1 on my 1D Mark II.  This causes the histogram on the back of the display to match fairly closely the histogram that I see later in ACR, my RAW conversion program of choice. 
[a href=\"index.php?act=findpost&pid=60408\"][{POST_SNAPBACK}][/a]

Eric,
It's late in the evening in Australia so I'll try tomorrow setting my in-camera image parameters to minus 2 contrast. Since I never shoot other than RAW, I've never bothered with such parameters. It never occurred to me that the way the histogram displays highlights would be affected by these parameters.

Just goes to show there's always something to learn. Thanks for that   .

[Ray]

Specifically, I work in manual mode with the in-camera spot meter on then point the camera at the brightest portion of the scene (usually in the sky), watching the exposure read-out in the finder and thumbing the shutter speed dial until the read-out says +3.0. That's my exposure, so I can now grab focus then compose the shot. This routine is simple and fast enough that I use it for all non-action shots.

Dale,
I'll try that technique too, tomorrow. If I've understood you, you are basically saying that the average sky should be over-exposed by 3 stops. A 'correctly' exposed sky (by the meter) would be a very solid, dark blue with every nuance of cloud detail visible, which would also, unfortunately, result in terribly noisy, under-exposed shadows elsewhere. It's the price we pay for a low DR camera.

Oops! +2 is the most I see in the 5D's finder, so I'll have to exercise the mathematical part of my brain   .

[Dale_Cotton]

Gary: nice explanation!

Ray:

> I'll try that technique too, tomorrow.

Good show!

> If I've understood you, you are basically saying that the average sky should be over-exposed by 3 stops

Now, now: I know you're just checking to see if I'm awake.  I'm saying that metering from the brightest part of the scene should underexpose a spot meter by X stops, where X happens to be 3 on my camera. If the brightest part of the scene happens to be the twinkle in your eye, so be it; if it happens to be that cloud over there shaped like a croc gobbling down a koala, so be it.

> Oops! +2 is the most I see in the 5D's finder, so I'll have to exercise the mathematical part of my brain

That's the problem when you buy one of those lousy budget cameras.  To get +3, just add 1 stop to the shutter speed reading at +2. Also: you may have +3.3 or +3.5 latitude on your 5D, since if I remember correctly you have more DR than my Pentax DS.

That said, what we're dealing with here is probably not as simple as half the full exposure latitude of your imager divided by two. It's the latitude from what your camera spot or centre-weight meters as the 0 point (middle grey) to the blow-out point. This depends on where Canon in its infinite and beneficent sagacity placed the mid point.

[Later addition: "brightest part of the scene" should actually read "brightest part of the scene that you want to retain detail". If something is specular or glaring white to my eye, I'm happy to let it blow in my capture.]

[madmanchan]

Eric,
It's late in the evening in Australia so I'll try tomorrow setting my in-camera image parameters to minus 2 contrast. Since I never shoot other than RAW, I've never bothered with such parameters. It never occurred to me that the way the histogram displays highlights would be affected by these parameters.

Just goes to show there's always something to learn. Thanks for that   .
[a href=\"index.php?act=findpost&pid=60421\"][{POST_SNAPBACK}][/a]

Hi Ray, no problem -- it's definitely not intuitive, because we're usually told that the in-camera parameters don't affect image quality if you shoot RAW.  That's true, in the sense that the raw data captured is unaffected.  But if you try to "optimize" your exposure using the histogram, as I do, then the in-camera parameters DO matter!  The reason is that the histogram is not based directly on the RAW data, but on the low-res JPEG image embedded in the RAW file.  The camera produced that embedded JPEG image using its internal RAW conversion routines and using the in-camera parameters, e.g. contrast, sharpness, color space, white balance, etc.

This means that things like contrast (e.g. -2 -1 0, 1, 2, etc.) and color space (sRGB vs. adobe RGB) affect the histogram a great deal.  

Eric

[John Camp]

I don't think you've got it quite right yet -- although you may (I may not understand what you're saying.)

Shooting RAW means that some in-camera parameters don't count for as much -- white balance, for example. If you note something white or a neutral grey in your shot, and can click on it later in Photoshop, then your specific in-camera white balance won't matter if you're shooting RAW. However, other camera parameters, like shutter speed and and lens stop, matter a great deal. If you underexpose or overexpose by six stops, shooting RAW, you can't adjust in Photoshop because the data is not there. If the data is not there, shooting RAW does not help. In other words, if you way overexpose, and flood the sensor with light, the RAW file will include lots of blown out areas, and the data that would be there with a proper exposure simply isn't there.

As far as the embedded .jpg goes, the fact that it's the .jpg being measured doesn't mean much. The histogram measures exposure (not sharpening) and what counts is the calibration (in your brain) of the relationship between the histogram and what you eventually see in RAW. If you expose properly, and have your in-camera calibration turned to high sharpen, then take your RAW file to Photoshop, the file will not be sharpened. It'll simply reflect an exposure. You can then sharpen as you wish. If your camera is set to RAW+.jpg (as some can be) and to high sharpen, then the RAW file will still be untouched, and the .jpg file will reflect sharpening. If the camera is set to .jpg only, then the file will be sharpened and to some extent, you're stuck with that. If, for example, you'd want to shoot a portrait of a young woman who happened to have some skin blemishes, and your camera was set .jpg-only, the blemishes would be emphasized and you'd have a harder time correcting that in Photoshop than if the exposure was more neutral, as in RAW.

Does that make sense?

JC

[Ray]

I don't think you've got it quite right yet -- although you may (I may not understand what you're saying.)

Shooting RAW means that some in-camera parameters don't count for as much -- white balance, for example. [a href=\"index.php?act=findpost&pid=60512\"][{POST_SNAPBACK}][/a]

John,
I think Eric has got it right. A problem with trying to expose to to the right results from the histogram not showing highlights that are slightly more than specral. That is, highlights which could be considered as specral but are in fact small patches of bright areas where the photographer really wants to preserve detail. (Detail in surf, foam and waterfalls, or bright patches of white feathers on birds and so on).

In other words, the histogram can be deceptive.

Even when shooting RAW, it seems that the histogram is based on an in-camera jpeg conversion, therefore, settings which apply only to jpeg images will affect the appearance of the histogram.

Having just changed the settings on the 'Picture Style' menu of my 5D to minimum contrast, I can confirm that the histogram now shows an exposure that is more to the right than it used to be, thus reducing the risk that in exposing to the right I might accidentally blow wanted detail in highlights.

[Jonathan Wienke]

A few remarks:

The in-camera histogram is derived from the in-camera JPEG conversion in every digital camera currently in production; I am not aware of any exceptions to this. It would be really nice if camera manufacturers would allow a histogram to be derived from the unprocessed RAW data, at least when shooting in RAW mode. It would actually be faster to calculate and require minimal firmware programming to implement. Given that, it is a good idea to experiment with your camera settings to find a setting set that causes the camera histogram and RAW data to match as closely as possible. On my 1D-MkII, that means color matrix 1 and daylight WB. The camera's natural white balance fairly closely approximates daylight. I initially had some problems with red channel clipping when shooting concerts indoors under stage lighting with a low color temp when using AWB, but after switching to daylight WB, the histogram clearly shows the red channel being farther to the right than the blue and green channels so I can back off exposure and keep it from clipping. The same principle applies when shooting around dusk or on heavily overcast days, except that the blue channel is the one most susceptible to clipping. In either case, using daylight WB allows me to see the unusual prominence of a color channel so that I can adjust exposure to avoid clipping, but since I'm shooting RAW, I still have complete freedom to set color balance as I see fit during RAW conversion.

It is a judgment call on the part of the photographer to decide how much of an image can be acceptably clipped. This varies greatly depending on subject matter. My personal philosophy is that in most cases, if an image element can be out of focus without detracting from the image, then it can be clipped to white or black as well, to approximately the same degree it is out of focus. If something is a foreground element and must be in sharp focus, it should not be clipped, either. But if it's in the background and heavily blurred, let it blow, let it blow, let it blow.

[CUclimber]

I just had to comment on how informative this thread has been to me.  I'm pretty handy with Photoshop and photography in general but I only got my first dSLR a few weeks ago and I'm still learning the ins and outs of proper exposure and workflow.  I've been playing with "exposing to the right" for a while now, but I can't wait to try some of this out tomorrow morning!

[bjanes]

A few remarks:

The in-camera histogram is derived from the in-camera JPEG conversion in every digital camera currently in production; I am not aware of any exceptions to this. It would be really nice if camera manufacturers would allow a histogram to be derived from the unprocessed RAW data, at least when shooting in RAW mode. It would actually be faster to calculate and require minimal firmware programming to implement. Given that, it is a good idea to experiment with your camera settings to find a setting set that causes the camera histogram and RAW data to match as closely as possible. On my 1D-MkII, that means color matrix 1 and daylight WB.
[{POST_SNAPBACK}][/a]

One way to get a raw histogram is to convert the file with DCRaw (do a google search) using the -m -n switches. Shown below is the histogram for the Nikon D200 with daylight exposure. The indicated patch (second from left, bottom) reads RGB 81, 154, 118. Since this is linear, the red channel is 1EV down compared to the green and the blue channel is down about 0.3EV. The Nikon sensor is not balanced for daylight.

Julia Borg has made an some interesting posts in the Nikon forum of DPReview.

The first post is about using a CC40M filter to balance the channels for daylight so as to improve dynamic range:

[a href=\"http://forums.dpreview.com/forums/read.asp?forum=1021&message=17142860]http://forums.dpreview.com/forums/read.asp...essage=17142860[/url]

The second is a custom white balance, which effectively turns off the white balance and allows the actual RGB values with gamma correction and tone curve to be displayed in the camera histogram.

http://forums.dpreview.com/forums/read.asp...essage=17723155

Bill Janes[attachment=344:attachment]

[jcarlin]

It would be really nice if camera manufacturers would allow a histogram to be derived from the unprocessed RAW data, at least when shooting in RAW mode. It would actually be faster to calculate and require minimal firmware programming to implement.

Jonathan,
    I don't think that this is in fact the case, well at least the bit about being faster to calculate.  Because in order to calculate a histogram (luminance or color) you have to perform interpolation for it to be accurate.  The compression is almost certainly "free" in any camera with a dedicated image processing chip (i.e. Digic X or Venus X for Canon or Panasonic) I doubt you'd be able to get histogram data any faster.  Furthermore since performance is something that every camera manufacturer lusts after if this where in fact faster I'd bet they'd do it, for all of the otherwise dumb things these guys do, I bet they do know exactly what they're doing when it comes to image processing.
    I agree that it would be nice to have the option of using the RAW file, but if one has a good idea of how a file is going to be processed, then it's equally nice to know what parts of the picture might get clipped during final processing (in a way I think this is like having your cake and eating it to).  This is something that just comes down to personal preference, and unfortunately we're all stuck with the personal preference of some engineer in Japan, just like mirror lockup on Canon cameras.

John

[digitaldog]

I think the first step is figuring out if the ISO setting correlates correctly with the meter. I take an incident reading. What I've done (and I could be totally off in this respect) is shoot (bracket) a white sheet of paper that is fully lit with sunlight. Then in ACR, I turn off all auto settings (Commend/Control U) and look at each with the info palette. I want the image that is as close to but NOT hitting 255/255/255/. At that point, I believe I've found the right bracket (expsoure) to get me pretty close to but not full highlight clipping. Obviously if you're using the in camera meter, it's fooled by white (thinks it's gray so compensate) but the idea is to see what gets me in that sweet spot exposure wise where white is real close to clipping. I would think a true specular would go over into clipping and I think that's OK.

I still have no idea how to sync up the on camera histogram since it's not based on anything I'm doing if shooting RAW (its the histogram of the rendered and encoded data). I'm not sure what we can do about that. I just ignore it. But at least I feel this gets me closer to moving to the right with respect to highlights. Does it seem rational? Seems to be working for me.

[Ben Rubinstein]

A couple of points I've mentioned here before:

For the most accurate histogram you have to take into account a few points.

Firstly your cameras WB needs to be close to the correct WB for a scene for the histogram to be rendered correctly. Shooting in tungsten light with a daylight WB will give a misleading histogram, the pictures once corrected will be 1-2 stops darker than the histogram may have indicated, and visa versa of course. If shooting in lighting which is nearer the edge of the spectrum, i.e. <4000K or 6000> then for a correct histogram it is vital to make sure that the WB is close to correct.

Secondly, although many people advise to turn the contrast/saturation, etc settings all the way down for a 'neutral' histogram, I disagree strongly. A histogram where there is '0' contrast and therefore retains the highlights is misleading unless you intend to print with '0' contrast!! Yes you have more room to play using the curves with the raw file, but the histogram is still misleading and you are causing yourself a lot more work when processing so as to add the contrast while holding back the highlights. I advise setting the jpgs to at least the lowest common denominator of your regular printing (admittedly far easier with people photography than landscape's), infact with many cameras you can set up multiple sets of paramaters for different occasions.

The idea is to let your histogram give you an idea of how much and what type of imformation you have captured, with the final print in mind. For that reason my pictures are all 'overexposed' on a laymans screen, but i shoot them that way, I keep in mind when shooting how the picture will look in print, i.e. with 'proof colours' selected in PS. Who cares how my pictures look on screen if the blacks block up in the prints I give to the clients.

Another thing to keep in mind of course is how your RAW converter renders the pictures with your usual paramaters. For example ACR shows my pictures about 1/2 a stop hotter in its histogram than does DPP/Zoombrowser. Yes the DPP histogram is exactly the same as my in camera histogram was, but I don't use DPP, I use ACR and I have to reflect that all the way back to when I'm taking the shot.

[BJL]

... I work in manual mode with the in-camera spot meter on then point the camera at the brightest portion of the scene (usually in the sky), watching the exposure read-out in the finder and thumbing the shutter speed dial until the read-out says +3.0. That's my exposure, so I can now grab focus then compose the shot. This routine is simple and fast enough that I use it for all non-action shots.[a href=\"index.php?act=findpost&pid=60411\"][{POST_SNAPBACK}][/a]

Dale,
   I too am inspired to experiment by your procedure!

My modified idea is to use spot metering on highlights with exposure compensation of about +3 in any of A, S or P mode. (Your routine sounds equivalent to using A mode with +3.) Then vary that compensation up and down a bit until I find the shift that works well for my E-1's combination of sensor highlight headroom and spot meter.

And then buy AEB insurance on the important shots, as Ray suggests.

[Dale_Cotton]

Glad to be of service.

use spot metering on highlights with exposure compensation of about +3 in any of A, S or P mode

You're 'way ahead of me! If this ever occurred to me I must have dismissed it due to my camera not permitting more than -/+2 stops compensation. Very impressive that your E-1 offers that. I'll be fascinated and envious to hear how it works out.