In-camera histogram, use and understanding
In the Camera forum we got off topic in discussing a camera comparison, but I wish to continue the discussion of histograms here.
Regarding the in-camera histogram I submit that there are way too many generalizations about the relative merits and value with regard to the accuracy of the histogram and how that relates to proper exposure.
What follows are my current views about the in-camera histogram and how I personally use the histogram in my own process.
I freely acknowledge that there is no current standard for either the histogram seen in a camera or in software. Manufacturers are free to interpret and to present exposure values in the histogram as they see fit. This means there is no single histogram.
That bears repeating; there is no single histogram.
What you see represented by your camera may not correlate well to what you see in different software and another camera may show a somewhat different histogram of the same scene. Even histograms from different camera models of the same manufacturer may not agree with each other.
Does that mean that a histogram has no value? Of course not. What it means is that the histogram is not a "quantitative" tool, rather it is a "qualitative indicator". It is, and must be, subject to interpretation. (Yes, I wish there were a histogram standard too.)
What I suggest everyone should do is to review critical images, good and bad, looking at the in-camera histogram and a histogram from trusted software. Adobe software has multiple histograms within its software programs and most of those agree with each other. If you will be using Adobe software, it makes sense to compare the histograms from your camera with the histograms within Adobe software.
I suggest that you compare shadow and highlight extremes as well as how colors relate to the histograms displayed. When you understand how different image data relates to both sets of histograms, you will also be able to use both histograms to their maximum benefit.
Fo my use, I divide histogram interpretation into the different capture situations:
1) RAW file capture, with obvious intent to post-process,
2) JPG file capture, again with the intent to post-process.
3) JPG capture, hoping to use the captures as final images.
In situations 1 and 2, I tend to "expose to the right", meaning that I want the highlights as far to the right in the histogram as I have previously determined to be safe. For most of my cameras that means that RAW captures I can often bump right against the right edge of the histogram in the camera, because I do observe some highlight recovery in post-processing. This is mostly valid for daylight color balance and a couple of my cameras are tested to need a little less than full right to be safe. For instance, with RAW capture my Canon 5D MKII with Highlight Tone Priority turned on is much safer to push to the right than either the XT/350D or 1D MKII cameras. The 40D is different still. My IR converted camera is greatly different too.
For situation #2, JPGs for post-processing, I tend to be more cautious and do not touch the right side of the histogram, especially with a lot of reds in the image. (Reds tend to saturate quickly and the histogram may not show red levels properly.) Artificial light is even more of a problem and I back off even more from the right edge because most digital sensors are designed for balanced sensitivity in the daylight spectrum. The skewed spectrum of tungsten light, for instance, can really be misunderstood in a histogram. Again, different cameras will show different histograms for the same situation so it's only important to understand "your" camera's histogram display.
For situation #3, JPGs to be used without processing, I tend to expose to put tones where they need to be according to a simplified "zone" system. If you have an 18% gray card, set your system to expose the card dead center in the histogram, if your testing indicates that is correct for your camera. You may have to adjust contrast to help control highlight clipping and shadow clogging. Different subject matter may require different strategies and compromises.
Again, I cannot stress enough that "you" need to test and know"your" camera's individual attributes, including the in-camera histogram. If you find that "shooting to the right edge" produces consistent over-exposure, that just means you have to back off a little. It is still almost always a benefit for post processing to shoot "towards the right", even in situation #2 JPGs.
Regarding the in-camera histogram I submit that there are way too many generalizations about the relative merits and value with regard to the accuracy of the histogram and how that relates to proper exposure.
What follows are my current views about the in-camera histogram and how I personally use the histogram in my own process.
I freely acknowledge that there is no current standard for either the histogram seen in a camera or in software. Manufacturers are free to interpret and to present exposure values in the histogram as they see fit. This means there is no single histogram.
That bears repeating; there is no single histogram.
What you see represented by your camera may not correlate well to what you see in different software and another camera may show a somewhat different histogram of the same scene. Even histograms from different camera models of the same manufacturer may not agree with each other.
Does that mean that a histogram has no value? Of course not. What it means is that the histogram is not a "quantitative" tool, rather it is a "qualitative indicator". It is, and must be, subject to interpretation. (Yes, I wish there were a histogram standard too.)
What I suggest everyone should do is to review critical images, good and bad, looking at the in-camera histogram and a histogram from trusted software. Adobe software has multiple histograms within its software programs and most of those agree with each other. If you will be using Adobe software, it makes sense to compare the histograms from your camera with the histograms within Adobe software.
I suggest that you compare shadow and highlight extremes as well as how colors relate to the histograms displayed. When you understand how different image data relates to both sets of histograms, you will also be able to use both histograms to their maximum benefit.
Fo my use, I divide histogram interpretation into the different capture situations:
1) RAW file capture, with obvious intent to post-process,
2) JPG file capture, again with the intent to post-process.
3) JPG capture, hoping to use the captures as final images.
In situations 1 and 2, I tend to "expose to the right", meaning that I want the highlights as far to the right in the histogram as I have previously determined to be safe. For most of my cameras that means that RAW captures I can often bump right against the right edge of the histogram in the camera, because I do observe some highlight recovery in post-processing. This is mostly valid for daylight color balance and a couple of my cameras are tested to need a little less than full right to be safe. For instance, with RAW capture my Canon 5D MKII with Highlight Tone Priority turned on is much safer to push to the right than either the XT/350D or 1D MKII cameras. The 40D is different still. My IR converted camera is greatly different too.
For situation #2, JPGs for post-processing, I tend to be more cautious and do not touch the right side of the histogram, especially with a lot of reds in the image. (Reds tend to saturate quickly and the histogram may not show red levels properly.) Artificial light is even more of a problem and I back off even more from the right edge because most digital sensors are designed for balanced sensitivity in the daylight spectrum. The skewed spectrum of tungsten light, for instance, can really be misunderstood in a histogram. Again, different cameras will show different histograms for the same situation so it's only important to understand "your" camera's histogram display.
For situation #3, JPGs to be used without processing, I tend to expose to put tones where they need to be according to a simplified "zone" system. If you have an 18% gray card, set your system to expose the card dead center in the histogram, if your testing indicates that is correct for your camera. You may have to adjust contrast to help control highlight clipping and shadow clogging. Different subject matter may require different strategies and compromises.
Again, I cannot stress enough that "you" need to test and know"your" camera's individual attributes, including the in-camera histogram. If you find that "shooting to the right edge" produces consistent over-exposure, that just means you have to back off a little. It is still almost always a benefit for post processing to shoot "towards the right", even in situation #2 JPGs.
0
Comments
I feel that some of the confusion my stem from what a person considers "overexposed" as opposed to another persons interpretation of it.
To me, overexposed means that you have blown the highlights. You pushed the exposure past the brightness latitude of the sensor. If it's truly overexposed, YOUR NOT GETTING ANYTHING BACK!
You CANNOT get back what's not there!!!
If you can recover data from what you thought was overexposed, then it was not overexposed, at least with concern to the sensor being able to record the data.
I also agree with Ziggy about getting to know YOUR histogram.
You can do it with software, or you can use a hand-held meter, I use both for testing purposes.
On my 5DMkII, the blinkies start right where they should, right when the histogram just touches the right edge. The only time this does not hold firm is when I have "Highlight Tone Priority" enabled. On my 5D2, I can push the exposure 2/3 stop before I blow the highlights. I've confirmed this with the hand-held meter and software afterwords.
BTW, exposing to the right (meaning NOT touching the right edge though) is VERY valid for JPG also. That right-hand fifth of the histogram holds a full 50% of your sensors data capability.
RAW & JPG: Use that right fifth, but stay just shy of the right edge.
Do some careful testing.
But if you can bring something back that's not there, I want you to be my financial adviser
There is one case where I can bring back what is not there. That is when the brighest part of the frame is monochromatic and fairly saturated and I only clip one channel.
The most common case is clipping the red channel and it happens because exposure meters are most sensitive in the green so they can miss red (or blue) highlights.
Clipping the red channel often results in color shift toward yellow because, as the scene gets brighter the green channel (which isn't clipped) catches up to the red (which is clipped). It commonly crops up in sunset shots and overexposed skin tones.
You can recover the blown out red channel by essentially cloning from the green channel in Photoshop. Here's how I do it:
1. If it isn't already, convert your image to 16 bit ProPhoto RGB. You'll want the dynamic range and wide gamut for this.
2. Go to Levels adjust the white output level to 128. This will give you the headroom you need for your new, brighter red channel.
3. Make a copy of your image in a new layer.
4. In the new layer apply the green channel to the red channel.
5. Go to blending options for your new layer and slide black blend if this layer slider up to to only show the highlights (say 100 or so).
6. Use the red curve adjustment on your new layer to match the red hue of your original image on your new layer.
7. Tweak the blend if and mask on your new layer to blend it into the background.
8. Flatten.
9. Use curves (and other tools) to adjust your image to taste.
10. Don't ask for financial advice.
Isn't one of the gotcha's on histograms when shooting raw, that the histogram is based on jpg settings and not directly related to the dynamic range that raw is capable of?
Scott Quier had a posting about how to set up your jpg settings so that the histogram reflects as closely as possible what raw is capable of.
http://www.dgrin.com/showthread.php?t=121623
http://www.danalphotos.com
http://www.pluralsight.com
http://twitter.com/d114
With all respect, as I highly respect your knowledge...
Didn't you just REPLACE data, not recover data that's not there
It wasn't until about a month ago that I learned how to actually read a histogram. Before that I chimped and usually ended up with underexposed images (need to calibrate the LCD on my camera to display the correct brightness).
The video that I watched (on the school), the guy, said to "expose to the right". Some people may misinterpret this saying. This does NOT mean to have it up and beyond the right wall.
No matter what you do, you want to be in the middle. You don't want to be against the right or left wall. You want it to gradually fade in and gradually fade away. here is an example of an underexposed histogram, a well exposed histogram, and an overexposed histogram.
When the histogram rides a wall, that means that there were shadows or highlights that were lost. Meaning, they are rendered as pure black and pure white.
The height of the bars does not matter. All that means is - how many pixels there are in the scene that are that shade of gray.
It does not matter whether you shoot raw or jpg. If you suck at exposing pictures correctly, shoot raw, it gives you more leeway to fix mistakes. I used to shoot raw, but I simply find myself exposing pictures correctly, and haven't had to make many adjustments in post so I shoot 100% jpg now. Saves time, memory, and money.
Actually, I take issue with these statement a little bit, Jeremy.
IF you are shooting in a snow field, you absolutely do NOT want your histogram in the middle - you do want much of your tones far to the right, even near the right extreme. You do want to be sure that each of the three channels, R,G,B, do not get fried, but you do not want them in the middle. If they are in the middle, your snow will look grey!
Like wise, you do not really want a black cat in coal mine to be entirely spread across your histogram either - although in this case overexposure - a bit - is better than under exposure, as you can pull you exposure slider back to the left a bit in RAW processing. If you are shooting jpgs, and your histogram is in the middle, your black cat will be rendered as a very grey one, not a black one.
The choice of RAW versus jpg is one of photographer's convenience and speed, versus finest image quality.
I do agree that exposure accuracy is far more critical for jpg shooters than for RAW shooters. To get good color accuracy with jpgs requires your exposure to be within 1/3 of an f stop in my opinion, If you think this level of accuracy is not needed when shooting jpgs, try shooting a Kodak grey scale with jpgs, and see if you can really capture every step of the entire black to white scale in your image.
Moderator of the Technique Forum and Finishing School on Dgrin
The candids may only require JPGs if it's a very inexpensive affair (i.e. I am not being paid very much) and I can't afford the time and trouble for processing. That's where controlling the ambient light can be a tremendous time saver and where proper use of flash plus modifier really counts. I try to limit my use of JPGs just because I like to keep the option of an adjustable WB and 16 bit processing, and that's only possible with RAW files. Mixed WB situations and difficult colors will make a RAW believer out of you in very little time and with a little practical experience.
I had some red carpeting in a church last year that I would not have been able to correct for to my satisfaction if I had shot in JPG. Red is not that unusual in a church but there was something unusual about this carpeting and I was surprised at how poorly the the camera handled it. Plus the red added a tint to everything, which I could adjust with a custom curve.
Do not underestimate the power of RAW capture.
Moderator of the Cameras and Accessories forums
Sports shooters do shoot jpgs almost routinely, and I would too in that venue. If I were shooting weddings, I would shoot RAW + jpgs these days. CF is cheap.
I always set my histogram to display all three channels, not just the luminosity channel. Too easy to fry red, as you mentioned above.
Moderator of the Technique Forum and Finishing School on Dgrin
As to RAW vs JPG... I'm not saying that jpg is as good as raw. its not. Does image quality really matter?? As long as you and, more importantly, your client are happy with the results, wouldn't making it more convenient for you be better? The same thing goes with blowing highlights... If you're doing it on purpose, and both you and the client like the result... who cares? Why get so technical about it? And the last time I checked... the client pays me the same whether I shoot raw or jpg...
If I'm presented with a situation where I KNOW it will be a challenge to pull it off in camera... sure, I'll shoot it in raw. Because I know that my talent isn't to a level that can handle that yet. But, can that situation be done correctly shooting jpg? You bet it can.
And I as well watch both luminosity and rgb histograms. Don't want to blow out the reds like everyone has said.
I just thought your comments about reading/using the histogram might be a little confusing to new users. Zone systems shooters understand the histogram intuitively, I believe.
Moderator of the Technique Forum and Finishing School on Dgrin
Yes and no. For parts of the image that are monochromatic, the red, green and blue channels are redundant; you only need one of them. When you convert a RAW image to RGB, the software is doing this for every pixel because each pixel only captures one color. Demosiacing (which is part of RAW processing) uses the neighboring pixels to guess at the true color of each pixel and sythesizes the other two channels.
In parts of the image where a channel is blown out, you really want to leave it out of the interpolation process. Different RAW converters handle highlight recovery differently and some are better about recovering chroma in the case where there is a single blown out channel than others. When I have a blown out channel in an important image I try several different converters to see which handles it best. However, if the blown out area is too large none of them will get it right so I do it by hand in Photoshop. If you are happier with the term highlight reconstruction than highlight recovery, I won't argue with you. However most of the software packages call this recovery regardless of what is going on under the hood.
At the end of the day ETTR metering is about capturing on the sensor what you need to create the image you want. I fairly often (mostly shooting sunsets) find cases where the highlights are relatively easy to reconstruct, so I deliberatly blow them out to capture more shadow detail.
All the rest of the peaks and valleys represent one of 256 shades of gray. The peaks and valleys will vary based on the range of gray tones in the image.
The very bottom line of the histogram represents one of each of the 256 shades of gray (pure black, gray values, pure white)
I really and truly wish that were true. Unfortunately there is no such thing as a single "histogram". Manufacturers are free to express the image data as they see fit. Since there is no histogram standard, the different camera makers are also free to interpret the data differently even by camera model within the same manufacturer.
As such the histogram representation of any camera can only be understood when compared to the histogram of an arbitrary standard. If you consistently use the same image processing software that has a competent histogram (a histogram that you can coordinate to and with actual numerical data readouts) and if you compare the camera's histogram to the software histogram, you can start to draw meaningful conclusions.
Once again I show these images, which show the histogram representations from several Canon cameras but the representation is from a single image file. If camera histograms were calibrated to the same standard you should see the same representation in the graphic display (since the data is identical for each), and yet you can easily see differences in the representations, meaning that none of the displays are necessarily accurate since they don't agree. (Similarity is not necessarily close enough for accuracy.)
The following are representations from several different software, compared to a camera's histogram. (The camera histogram is the middle image on the left column. Note too that this is from a different file from that used above.)
Again, my recommendation is to choose a single software product and makes comparisons with its histogram and your camera's histogram. You will see some correlations and then understand better what the histogram of the camera is really telling you and how you need to interpret the boundaries as well.
Moderator of the Cameras and Accessories forums