Histogram and Dynamic Range

Inquiring minds would like to know stuff. Answer if you are able and comment if ya like.

Firstly is this statement close to accurate? The histogram shows the amounts of digital information in the image and which areas have more or less of the information.

I have been underexposing to avoid blowing out the highlights. In very good lighting this has worked well but in less than favorable lighting the results are like turd at the dark end. Very noisy at the dark end, beyond repair kind of noisy at times.

Zooming into the dark end on a "peaks to the left" image easily shows the noise. . . . Zooming into the highlights on a "peaks to the right" image does not show much noise at all.

Is this a correct statement? Noise in a digital image is not part of the scene captured.

Does the histogram peaks to the left include the additional information of the noise, making the peak taller there? If so, is their less digital information in shadows than there is in highlights? Is there more digital information in the highlights, thus preventing the noise to infiltrate?

Is there correlation of dynamic range and the histogram we see in lightroom?

Now I am even dizzier from composing this.

This article on Cambridge in Color might help answer some of your questions: Digital Exposure Techniques.
....

Bourbon Neat said:

Inquiring minds would like to know stuff. Answer if you are able and comment if ya like.

Firstly is this statement close to accurate? The histogram shows the amounts of digital information in the image and which areas have more or less of the information.

I have been underexposing to avoid blowing out the highlights. In very good lighting this has worked well but in less than favorable lighting the results are like turd at the dark end. Very noisy at the dark end, beyond repair kind of noisy at times.

Zooming into the dark end on a "peaks to the left" image easily shows the noise. . . . Zooming into the highlights on a "peaks to the right" image does not show much noise at all.

Is this a correct statement? Noise in a digital image is not part of the scene captured.

Does the histogram peaks to the left include the additional information of the noise, making the peak taller there? If so, is their less digital information in shadows than there is in highlights? Is there more digital information in the highlights, thus preventing the noise to infiltrate?

Is there correlation of dynamic range and the histogram we see in lightroom?

Now I am even dizzier from composing this.

Click to expand...

Your histogram shows the tonal variation of your image as it is being displayed. It doesn't necessarily say much about the actual information it (the shot) contains. If you underexpose in post, your histogram adjusts to that while the RAW still contains the same information.

There is no difference between a pixel value and the noise it contains. Noise is a part of the value. We talk about noise but it doesn't exist as something independent of the signal. The signal the sensor receives is never 100% accurate and when it is processed by the cam, there is more "noise" (inaccuracy) added.

Underexposing results into more noise since the accuracy of the signal is directly related to the size of signal captured. By lowering the light that reaches the sensor, you lower the signal and thus increase the noise. It is more visible in the darker parts of the shot because the lack of light is what makes pixels dark to begin with. Their captured signal is low and thus noisier.

To add; your dynamic range has little to do with the histogram. It's the difference between the darkest and lightest part of your shot. You can affect the captured range in post but the histogram itself isn't an accurate representation of it.

Horoscope Fish said:

This article on Cambridge in Color might help answer some of your questions: Digital Exposure Techniques.
....

Click to expand...

Thanks for the directions. An hour and a half later I finally forced myself to leave. A lot of good information at that website.

J-see said:

To add; your dynamic range has little to do with the histogram. It's the difference between the darkest and lightest part of your shot. You can affect the captured range in post but the histogram itself isn't an accurate representation of it.

Click to expand...

The histogram expresses the tonal range and the magnitude at various locations in the range. Is that correct?

Is the histogram stop based from one end or the other?

Bourbon Neat said:

Does the histogram peaks to the left include the additional information of the noise, making the peak taller there?

Click to expand...

The histogram expresses the tonal range and the magnitude at various locations in the range. Is that correct?

Click to expand...

These statements are not correct. The histogram does not show magnitude.. period.

The histogram is a simple bar chart that shows the pixel count of all the brightness values from 0 to 255. The height is simply the count of pixels with that tonal value. Any peak is only there because there are a lot of pixels with that same value. The peak might be about dark or bright values.

If your image is mostly dark, there will be many dark pixels, so the the peaks will be at the low half.

If your image is pretty bright, there will be many bright pixels, so the peaks will be in the bright half.

There is no right answer about how it should look, since it depends on the scene in front of the camera. But dynamic range implies it should be fairly well filled, with some dark and some bright values. A very bright color should approach 255, but if the scene has no bright colors, then maybe not. The reason we pay attention is because anything past 255 is clipped, lost, not recoverable. A tall peak right at 255 is a count of many clipped pixels.

The vertical scale is always scaled, so the tallest peak will always be about full scale. Height has no absolute meaning, everything is relative. Because the histogram simply shows pixel count of each of the various tone values. It simply shows how the brightness range of your image is distributed, from dark to bright.

In the camera, there is one single gray histogram (technically called luminance, which is just a math model to represent gray scale values, and it cannot show clipping. It is NOT real data).

And there are also three individual Red and Green and Blue channel histograms, which do show the images real genuine non-math-manipulated true values. Ignore the single gray one, and for examining clipping purposes, pay attention only to the three RGB histograms. See There are Two Different types of Histograms for more.

And for what it's worth, the histogram is also showing the gamma encode values (the RGB data is gamma encoded). This means the middle gray midpoint is NOT at 128 mid-scale. It would be for linear data, but instead, the gamma value for that is about 187, or about 3/4 scale. However, this value is a bit vague, because the camera white balance and contrast and color profile setting shift it somewhat, in a variable way.

Bourbon Neat said:

The histogram expresses the tonal range and the magnitude at various locations in the range. Is that correct?

Is the histogram stop based from one end or the other?

Click to expand...

The histogram is just a visual tool to give you some rough indication of the current situation but it's never exact science.

If you look at the both opposite ends of a histogram, you notice it ranges from 0 to 255 or black to white. That's already a very compressed measure to express your tones and dynamic range. Dynamic range is expressed in exposure stops which means that the next stop is always twice as "bright" as the previous. If you have a dynamic range of 10 stops, the highest stop is 1024 times brighter than the first which implies there are 1024 different values between both ends of the spectrum. That exceeds what the histogram can show.

That most monitors only show an 8-bit variation has more to do with it than the impossibility to represent it fully. It simply displays what you see.

I haven't spent the time to read through all of the responses thoroughly, but I offer simply this. The histogram shows the range of light information captured in your image on a scale of 0 (black) to 255 (white). Anything at the extremes that is not meant to be either black or white (ie. 'spikes') represents potentially lost information, so if you have a spike you need to adjust appropriately, and if you have two spikes then you have more dynamic range in the image than your camera is capable of capturing (which is where HDR techniques become necessary).

There's a theory that I adhere to that recommends you meter in a way that produces your bump to the right of the histogram and not the left. Recovering dark regions from a well exposed (no spikes) but bright image will generally produce less noise than brightening (amplifying) shadows from a well exposed but dark image. "Shoot right" is the catch phrase in case you ever read it.

A histogram is a tool, and like a hammer you can build a box with it, and you can build a cathedral with it - it's all in knowing how and when to use it (and ignore it). I've seen amazingly complex color correction work done based on the histogram and where the RGB spikes are, and if you're into nailing your image in every way possible then this is where art and science collide. But for the average photographer I'd say a quick glance at it with your shot on the back of the camera is generally sufficient once you understand what you're looking at.

Thank you to all who responded, your knowledge is very helpful.

BackdoorHippie said:

There's a theory that I adhere to that recommends you meter in a way that produces your bump to the right of the histogram and not the left. Recovering dark regions from a well exposed (no spikes) but bright image will generally produce less noise than brightening (amplifying) shadows from a well exposed but dark image. "Shoot right" is the catch phrase in case you ever read it.

A histogram is a tool, and like a hammer you can build a box with it, and you can build a cathedral with it - it's all in knowing how and when to use it (and ignore it). I've seen amazingly complex color correction work done based on the histogram and where the RGB spikes are, and if you're into nailing your image in every way possible then this is where art and science collide. But for the average photographer I'd say a quick glance at it with your shot on the back of the camera is generally sufficient once you understand what you're looking at.

Click to expand...

Yes, it is an excellent tool and have every intention of learning to work with it. The theory I had been using worked fairly well and failed very well too. Noisy darks. Your theory of having bump to the right is making better sense to me after more research. Some folks refer to it as ETTR. The turth is, I had not really paid enough attention to the histogram at all, in the camera not at all. Today is a new day in that regard.

Bourbon Neat said:

Yes, it is an excellent tool and have every intention of learning to work with it. The theory I had been using worked fairly well and failed very well too. Noisy darks. Your theory of having bump to the right is making better sense to me after more research. Some folks refer to it as ETTR. The turth is, I had not really paid enough attention to the histogram at all, in the camera not at all. Today is a new day in that regard.

Click to expand...

I abide by ETTR and find it works very well. The main thing is to avoid blow-outs because once you've blown-out, whether it be in the highlights, shadows or an individual color-channel, that's it... Detail in that range has been lost and there's no getting it back.

I don't mean to beat a dead horse, but at the same time I don't feel that point always gets the emphasis it deserves when discussing ETTR and histograms in general.
.....