Histograms

[Colin Southern]

While we are talking about histograms, no one has mentioned the RGB histograms one gets with live view on the latest Canons. One can switch between the levels histo and the RGB histos. This presents a new problem for me because I meter manually a great deal of the time but the advice is that, if one is shooting in raw, the prominant color should be firmly nudged to the right. If however, one has metered on something away from the real focal area this action may not be possble. So I have been ignoring the RGB histos. Should I? Or am I wasting a resource?
Paul

Hi Paul,

The RGB histogram is handy when you've got fairly extreme colour temperatures - and you can see how close you are to blowing particular channels. You can get these kinds of temperature shifts shooting under certain types of lighting and when shooting the likes of sunrises - sometimes you can end up - for example - with the blue channel leading the red channel by a full couple of stops (or vice versa).

In terms of where to place the histogram, it really depends on the scene that your shooting (see my earlier post in this thread). In GENERAL terms you usually won't want areas of blown highlight (unless they're the likes of specular reflections or unavoidable back lighting) and you usually won't want to under-expose a shot by too much (in terms of how it looks on the histogram, not how it looks on the review screen) as you'll be throwing away levels and end up with shadow noise and posterisation when you correct in PP. The exception is particularly flat scenes where - if you expose too far to the right - it's difficult to get them looking realistic, I assume due to the vaugrancies of gamma and tone-mapping, but I really haven't thought that part through too much.

The likes of spot metering is great because it allows you to put tones where they should be, regardless of any other crazy lighting that may be present in the scene that may upset automatic metering modes.

[Davey]

This might come across wrong or stupidly simple but what I'd recommend is just forget all the technicalities of what is what since it seems that kind of approach isn't your thing. Some people pick up technical details without any help and some don't, I've often seen people just get more and more confused when someone who gets it tries to explain it to someone who doesn't so I'd say forget it, at least for now.

I don't want this to come across wrong but you might struggle with that choice of cam for a switch over to digital. Problem is that camera isn't really one to jump into digital with as a learner since it's not film and although artistically the same rules might apply physically it's a whole different world and fundamentals like how exposure works is very different. Not so steep a learning curve if you are the kind of person who can figure out exactly what the histogram is alone without anyone telling you or reading the manual or articles on it. I don't own a D3 (I wish hehehe) but know 2 people who do, one a high level enthusiast with need for more control than average amateur needs and the other a pro who mainly shoots architectural work for comission. That's the level of user it's aimed at, as a result it has a lot of features which you have a lot of manual control over and hence need to understand how they work to take a picture.

With those things in mind you're going to have a steep learning curve, steeper since coming from film you even have some knowledge you need to unlearn and expectations based on film experience which may hinder you more than help as your digi will never behave like your old film camera. Even if you can get a similar result (or maybe even identical depending what you do) out of it thats gonna take a completely different method and experience to acheive. Same destination but 2 completely different paths if you like. I don't want to sound patronising or teach you to suck eggs but I'm guessing digital being completely new there is a lot of stuff you won't automatically know without first hand experience and some stuff you could be a master at on film formats that you'll need to learn to do from the ground up and completely different way.

So my advise would be forget the books and technicalities, just try experimenting with it, trial and error is free (time and battery charging aside) with digital unlike film. Fill a mem card (or two if you're using both slots) and you can just blank it and start again no worries. Figure out basic stuff 1 feature at a time to avoid complexity and forget about if the image is good or looks like your old film cam result. For instance with histograms keep it simple and take a few pics in an evenly lit room of varying subjects but with same fstop/shutter speed. If you rig the scene so it's just dark things, then rerig for just pale things and midtones and then a mix. Now look at your histograms for each image. Little things like that will probably be helpful in figuring it out the untechnical simplified fact is in essence it's just a measure of all the stuff in a scene and how dark or bright that stuff is.

Then when you experiment more you can introduce real world applications of it and related features like figuring out what happens if too much light is detected eg. something like set it to too long an exposure and snap bright stuff with details like very pale grey details on white etc etc. Then you can see on the image anything that has exceeded what the digital sensor could measure is not accounted for and the brightest white possible exists in it's place. Then when you've tried combinations of this and are familiar you and see this too far out of range over exposed stuff shows furthest rght on the histogram and flashes on the image you can then learn the technical name is "clipping" and terms like "blinkies" or exposed to far right etc etc make sense because you can relate them to simple things you've actually experienced and observed so it's not a dry mental exercise. Sorry it's a bit long winded but I hope this helps and didn't know how to explain in less words without terminology that might cause more confusion than help. You get there soon enough, and I look forward to seeing your pics. Good luck

[Colin Southern]

See my 1st post in this thread Colin, No 8. You can leave the 'replay' at 'Highlights' or 'Histogram' (or a few others which I don't use, mostly file data).

Hi Chris,

That was the bit that I was wondering about - I wasn't sure if you meant it was something that had to be turned on on a case by case basis when reviewing historical images, or if is was something that could be turned on permanantly and show when reviewing the image immediately (and automatically) right after capture.

With Canons you can set an automatic review time and have both Histogram and Blinkies displaying simultaniously - was just assuming that Nikons could be setup the same?

[Mossy]

Depends on the model AFAIK.

My D90 can do highlights OR Histogram OR Channel histograms. Not 100% sure on D3.

[peter2108]

...

1. If you're shooting outside on a sunny day - sun is behind you - and your shooting at your local park, then an ideal exposure would be one where the histogram gently kisses the right hand side of the scale ... even if the image looks quite over-exposed on the screen (we bring this back in post-processing) (the over-exposure look is still desireable so long as there aren't sharp peaks at the end of the histogram indicating areas of blown highlights, as this means that the maximum amount of shadow detail was captured with the minimum of noise) (ie best signal to noise ratio).

If I set my Canon 400D on M(anual) I can set the aperture and then change exposure time by twirling the wheel at the front. This moves the exposure level indicator as the aperture/exposure time combination becomes more or less over/under exposed.

If I understand Colin's rule I can push this well to the right past the central correctly exposed reading. Then take the picture and check the histogram.

So really I only need M and the automatic settings?

Peter

[Paul]

Colin, I love the colors in the picture you posted in #22. Particularly the tone of the greens. How much PP did you do, if any?

Paul

[perth45]

Chris,
many thanks for doing that....no I didn't go inside, I didn't realize you could, but I'd love to especially now you've raised my awareness of it........The picture was only altered in contrast really, which improved it somewhat.....I'm off to Salford this weekend to play around with my camera and check out whats happening with the histograms......
Davey,
good advice from you also... I think you are so right....many thanks.....

[perth45]

Colin,
Thanks for that information and all the help....I really appreciate it from everyone......The photograph in #22 was taken on my mates D40 which I borrowed and played with one weekend.... I used a polarizer only, then warmed it up in camera....again....if I'd taken it on Fuji Velvia I would have added an 81C warm -up and a grey grad to add depth and mood to the sky....I'm sure I can do all this using photoshop but thats not really the direction I want to go in.....since buying my D3 I've bought a Polarizer (b&w) and i'll experiment with my tried and trusted filters to create the image in camera....however...would I be correct in saying that the quality of the image may be diminished because of the filters, therefore the answer is STILL to photoshop it.....What do you all think?...and just out of curiosity...if I took the same image on IS0 200 and then again on say, 1600....what difference does this make to what I see on the histogram, or would it be the same pattern ?........

[perth45]

Chris,
Diverting a bit but have you got any interior shots???????????

[Colin Southern]

If I set my Canon 400D on M(anual) I can set the aperture and then change exposure time by twirling the wheel at the front. This moves the exposure level indicator as the aperture/exposure time combination becomes more or less over/under exposed.

If I understand Colin's rule I can push this well to the right past the central correctly exposed reading. Then take the picture and check the histogram.

So really I only need M and the automatic settings?

Peter

Right track in theory, but in practice you can only push some scenes so far before it becomes very difficult to adjust the exposure in PP to the point where it looks natural.

Eg if your metering a very white scene (eg mostly snow) the camera will see gray and under expose it by around 2 stops - so if you put the indicator 2 stops to the right you may be close histogram wise, but if the scene is the proverbial "black cat on a black rug" then the camera will see gray - it will inherantly over-expose anyway - and putting the metering 2 stops to the right will probably mean your up to 4 stops over - still good histogram wise - still good in terms of ideal signal to noise ratio, but when you expose what is esentially a shadow as a highlight you then have a bit job to return it to a shadow in PP and still have the gamma and contrast look good.

But for less extreme scenes your pretty much on track. Essentially you can shoot any scene in manual mode - but it you're just "out and about" doing a bit of fun shooting then it's probably easier to just jet the camera do it's thing in the likes of Av mode. Even when I'm out shooting professional landscape I'll still use Av mode a bit. With constantly changing light it's easier to use Av mode with X amount of EC dialed in so that it gets the Histogram where I want it (and no blinkies) from there on in the light will change rapidly, but the camera compensates automatically - so I only have to make minor tweaks to the EC until I get to 30 sec shutter speeds - then it's either up the ISO, open up the aperture, or go bulb mode (aka "super manual") for shutter speeds longer than 30 seconds.

Make sense?

[Colin Southern]

Colin, I love the colors in the picture you posted in #22. Particularly the tone of the greens. How much PP did you do, if any?

Paul

Hi Paul,

My apologies - I seem to have mislead a bit here

I was trying to show Greg how easy it is to post an image inline - so I went to his FlickR gallery and picked one of his as an example.

So it's Greg's shot, not mine - sorry for the confusion!

[Colin Southern]

Colin,
Thanks for that information and all the help

No worries

would I be correct in saying that the quality of the image may be diminished because of the filters, therefore the answer is STILL to photoshop it

The more you discuss this the more you'll probably get "polarised" (no pun intended) opinions from the "purists" -v- the "practicalists" (if there is such a word). Many will argue that any additional medium in front of a lens will (in theory) degrade an image - but in reality this "degradation" is undetectable. For me, the bottom line is that filters are there to do a job - so I use them if I need them to do that job.

Some like to get things right "in camera" (myself included) - some like to do it in photoshop. In reality though, most of us shooting at a professional level are probably doing both; my approach is to get it right in camera as best I can (especially when using the likes of GND filters) - and then make whatever subsequent adjustments are necessary in PP.

Often I'll hear people say "it's just easier and cheaper to do it in Photoshop" but it's my observation that if your shooting professionally then it takes pretty much the same degree of skill to get it right "in camera" or "in Photoshop" - it's one thing to do a quick "hatchet job" on an image that only ever displays here 680 pixels wide (and looks good) - it's another thing entirely to work on an image like most of mine that are printed out 44 inches wide.

At the end of the day it's all about a co-ordinated system of processes - it starts with the capture in camera - it passes through the post-processing stage - and it finishes with printing (and in my case over-spraying and stretch mounting). All require skill. The photographers who's work I respect the most are very competant photographers - but also very competant in Photoshop.

Just out of curiosity...if I took the same image on IS0 200 and then again on say, 1600....what difference does this make to what I see on the histogram, or would it be the same pattern ?........

Assuming that you adjusted your shutterspeed and/or aperture to compensate then in theory yes, but in practice the dynamic range (range of brightnesses) of the sensor decreases as the ISO increases (usually by around 3 stops at the high end of the camera's ISO range) - and this will have an effect on the histogram. As a rule of thumb, if you're tring to capture a very contrasty scene then you'll always be better off using a lower ISO.

Does this help?

[Dave Humphries]

...and just out of curiosity...if I took the same image on IS0 200 and then again on say, 1600....what difference does this make to what I see on the histogram, or would it be the same pattern ?........

Hi Greg,

Well, assuming you also adjusted the shutter speed and/or aperture to get the correct exposure, I don't think there would be an appreciable difference on the histogram, but the resulting photgraph would show increased noise. Although on a D3, this shouldn't be as bad as many cameras.

I couldn't find sample D3 images, but here's a link to the image samples page of Steve's Digicams review of the D700. If you look closely at the images DSC_0328 and DSC_0347 (Click to open them), they show the same test shot at 200 and 1600 ISO.

Hope that helps,

[alextor]

I am also completely new with histograms.

I just read http://www.luminous-landscape.com/tu...se-right.shtml, as a poster kindly suggested. It was helpful.

My question is, what is the point of exposing to the right if it will later be necessarily to decrease the exposure of the image in post processing to achieve a correct exposure?

According to the LL article, the benefit of exposing to the right is that the brightest 5th of the histogram contains half of the exposure values of the dynamic range. So, if we expose to the right, we will have more nuanced colors.

But, if we later want to decrease the exposure for the finished product, won't we be losing all of those nuances as the image decreases in exposure?

Perhaps the confusion stems from the linear/non-linear distinction. I get that the sensor is linear, whereas the eye is not. Because the sensor is linear, RAW files must also be linear. But, what about JPEGs? They must not be linear or else none of this would make very much sense to me. Could anyone please go into some detail about RAW to JPEG conversion, and how it benefits from exposing to the right? I assume that it is the key to the method.

Thank you all for sharing your knowledge.

As a side note, it certainly makes sense that exposing to the right would decrease noise in the final product. However, this seems separate from the nuances issue. Is it?

[Colin Southern]

Hi Alex,

I've only got a few minutes, so I'll answer briefly now and expand on it a bit later for you if you need me to.

In a nutshell, the further the histogram is away from the right, the more potential levels your throwing away, which takes what's left closer to the noise floor.

If you under-expose by 2 stops (which is barely noticeable to the naked eye) (still linear at this stage) you've just thrown away a full THREE QUARTERS of the possible information that you could have captured ... the net result is what's left gets stretched over a wider range (to the point where the dirrerences between levels may become visable), and you're also having to increase the level of the shadow areas which are (at this point) VERY close neighbours to the noise floor.

So (with a few exceptions that I can point you to a bit later) not shooting to the right opens up the potential for noisy shadows and postorisation.

Whether or not you have to subsequently reduce the exposure in post-processing depends on how contrasty the scene is - if it's quite flat then you will have to reduce it (and large reductions can be difficult to get looking natural), but more often than not the scene will look better right out of the camera (ie without being knocked back) anyway (if you don't expose to the right then chances are you'd be increasing the exposure in post-processing).

Does this help?

Happy to expand on this some more later if you need it.

[crisscross]

Point 1: in the camera replay window look at both the histogram and highlights together; just a few highlight spots are appropriate for white paintwork and birds in bright sun and odd cloud fringe - then return to histo & it may well be nuzzling the right hand edge

Point 2: shooting RAW/NEF the camera has preserved info 10-20% beyond the edge of the histo as it arrives in PP prog, so you can 're-centre' it with exposure compensation AND (less obviously) colour temperature.

Point 3: if in the habit of taking impossible landscape shots with horrid haze etc etc, what plops into the opening window can look pretty unpromising (see my NX2 gallery on pbase). BUT there is a wonderful moment when by various jiggling and squiggling you correct the exposure of at least a zone of the picture and suddenly the saturation has improved as well as the luminosity. Don't ever forget the camera is a machine, smart on the light metering, but totally dumb as to what you are pointing at and why. Only you can tease that out.

[alextor]

Colin (or anyone, if they have the patience , to help me clarify, could you run through the following scenario with me?

Say there is a scene that normally you would capture with most of the pixels around the middle of the histogram. According to the "expose to the right" rule, it would be better to shoot the scene so that the pixels kiss the right side. Let's say that in this scenario, this brightens the image, and puts more of the pixels in the highlight region.

In post processing, you look at the picture and you think that it is overexposed. You want it to be darker to convey the original mood. So, you pull it back to the exposure you would have normally shot. Is it not true that all of the nuances gained from exposing to the will disappear, and posterization will occur upon pulling the exposure back in PP?

This is based on the assumption that the output format (say JPEG) is also linear.

In my mind, in this scenario, exposing to the right adds nothing in terms of the nuances of color after post-processing.

To summarize, I understand that exposing to the right increases the signal to noise ratio and allows for more nuanced colors if you keep the image exposed to the right in post. If you scale it back in post, the signal to noise ratio will still help your picture, but won't the benefit of the gained color nuances disappear?

[Colin Southern]

Say there is a scene that normally you would capture with most of the pixels around the middle of the histogram. According to the "expose to the right" rule, it would be better to shoot the scene so that the pixels kiss the right side. Let's say that in this scenario, this brightens the image, and puts more of the pixels in the highlight region.

If there were such a scene. This would be a scene with few highlights, few shadows, and essentially all mid-tone, which would be quite unusually flat. Far more common to have both shadow and mid-tone areas, at a minimum.

In post processing, you look at the picture and you think that it is overexposed. You want it to be darker to convey the original mood. So, you pull it back to the exposure you would have normally shot. Is it not true that all of the nuances gained from exposing to the will disappear, and posterization will occur upon pulling the exposure back in PP?

Yes - but they'll be cleaner nuances. What your suggesting is fine in theory, but it's based on the assumption that essentially whats coming out of the camera is already what your wanting in the finished photo, which (for me anyway) is rarely the case; by under-exposing (in terms of the sensor's potential) your losing headroom / safety margin - and the problems will start when if you subsequently discover black shadow regions that you want to recover detail from. Perhaps a good example is in these two shots ...





The first was processed in CS2 - with the black clipping point set to 5, and you'll see that I've completely lost all detail in her black jersey. The second shot was processed with CS3 and I've used the Fill Light control to recover the detail. These were both shot in RAW with fairly optimal histograms; if I hadn't "exposed to the right" there would have been a LOT more noise in that 2nd shot (as it is it was still quite noisy, but I got away with it).

This is based on the assumption that the output format (say JPEG) is also linear.

Not sure what your meaning by a linear JPEG output format - once it's been through gamma conversion it's not linear, and it'll always have been through gamma conversion by the time it's written to a JPEG file.

In my mind, in this scenario, exposing to the right adds nothing in terms of the nuances of color after post-processing.

If you get it exactly right in camera then any potential gains may not be visable. Keep in mind though that sensors are working in a LINEAR response "curve" whereas the histogram your seeing is already gamma converted (based on an in-camera generated JPEG, even if your shooting RAW). If you want to see how easy it is to under expose 2 or even 3 shots, stop your lens down 2 or 3 stops from wide open and look through the viewvinder whist repeatedly pressing and releasing your depth of field preview button. You'll probably be quite shocked at how small the difference is - the yet a full 3/4 (2 stop) or 7/8 (3 stop) of your sensors capturing ability is in this region ... so I'm sure you can imaging how much ISN'T left when you get down to the shadow regions (in linear gamma).

Making sense?

Having just said all that, with extremely flat scenes you can also get a situation where exposing to the right can bite you in the rear end; your likely to find that t's difficult to get a natural looking result if you have to drag the exposure a long way down. I haven't thought it through too much, but I suspect it's something to do with the gamma curve used by the computer not matching what our eyes are used to. Specular highlights / back lighting are another case where you may want to actually over-expose parts of the image (spiking the histogram at the right).

[Dirk]

Hi, I felt just like the TS, knew how to get a good looking photograph out of my Nikon F65, then I went to digital with the first Canon EOS 300D with what later seemed a hacked firmware , it worked but not the way I wanted it, and sold it on a market place here in Holland for a little over the money it cost me

I went to work a day later and looked just before leaving on the internet and saw the Nikon D70, bought it in lunchtime with only 1995 clicks and a 28-80 in "kit" second hand.

The first impressions where aweful, I had still the impression that it was the huge amount of menu options that killed me at that particullairy moment , now about 2 years later and almost 4000 images later I have the feeling I am getting to know it

Well I have assembled a lot of photogtaphic informattion I am likely to share :

http://photocamel.com/forum/nikon-fo...knowledge.html

http://www.luminous-landscape.com/tu...stograms.shtml

http://www.eventsinlife.org/index.ph...=descriptions2

http://www.bol.com/nl/p/boeken-engel...384/index.html

or also a good book

Encyclopedia of digital photography by tim daly