Would love a histogram lesson

4labs

I would really apreciate a quick tutorial on reading a histogram or a good reference source if noone wants to bother. I have been using manual exposure since Yoemite and for the most part am happy with my results but sometimes I am so off that I would love to be able to read a histogram better to correct my errors.

Thnxs in advance.

patch29

Start here. There are a lot of good tutorials at the Luminous Landscape.

4labs

Thnxs Patch!

patch29

4labs wrote:

Thnxs Patch!

No problem, here are two more articles from Digital Outback Photo, One and two.

Giphsub

Great references guys I understood about where the histogram needed to be and what it needed to look like, but now I know why!

wxwax

I my histogram.

rustic

wxwax wrote:

I my histogram.

I read somewhere that the histogram on my Nikon D50 only reflects the green channel, and thus isn't necessarily that useful. Does anyone know if that is true or not, and if so, if there any sort of workaround (short of upgrading:D)?

I still use it and haven't noticed any problems, but I'm also not necessarily at the point where I would notice subtle problems even if they were there.

Thanks!

wxwax

rustic wrote:

I read somewhere that the histogram on my Nikon D50 only reflects the green channel, and thus isn't necessarily that useful. Does anyone know if that is true or not, and if so, if there any sort of workaround (short of upgrading:D)?

I still use it and haven't noticed any problems, but I'm also not necessarily at the point where I would notice subtle problems even if they were there.

Thanks!

Interesting. I see the same claim here, and a workaround (use the blinkies) to avoid over exposure.

I've shot in clubs with lots of red and no other light, and had histograms deceive me. Not many cameras have histos which show all three channels, which is what you need in these kinds of lighting environments. A histo which only shows one channel, or is a combo of all three, might not show that one of the colors is overexposed. So I guess a combination of common sense, the meter and the blinkies will have to save the day (or night.)

ziggy53

I just ran a test against 1 image using the histogram of the Canon dRebel XT/350D against several software histograms:

Canon's DPP (ver 2.2.0.1)
Adobe PSCS2 (and PS 7)
GIMP 2.4 (just released)
Picasa (ver 2.7)
PhotoEditor (ver 1.0 I think)
Magic Enhancer lite (inside of IrfanView)

While none, I repeat none, of the histograms agree perfectly, Canon's DPP came close to the XT histogram with the following exception, the red and green middletones are skewed right a bit on DPP versus the XT. Otherwise, the camera histogram does seem to indicate a full RGB palette. The highlights and shadow regions on the camera appear to be very similar to the full RGB on DPP.

Photoshop versions did not matter, they displayed the same, just not the same as DPP.

PhotoEditor was also fairly close to the camera histogram in shape. Next closest was Picasa.

The point, if there is one, is that there is no such thing as a single histogram interpretation. I had previously determined that DPP was a pretty good match for the camera, and this test proves it to be so, with some exceptions. Get to know what the histogram means on "your" camera and software and you will be able to use it properly and accurately. The highlight overexposure indicators, plus the histogram, plus spot metering is an incredible arsenal of information that many modern dSLRs have. Use the tools you have correctly and accurate exposures should not be a problem.

As I get time, I will do some more tests, but this will take weeks because I am in the middle of a large "paying" job. (It doesn't pay that much but it is a large job just the same.)

Here is a single image showing the results. Note that the XT display was elongated horizontally to provide an easier visualization and that the DPP display is squashed vertically for the same reason.

P.S. the camera histogram is the middle image on the left column.

pathfinder

Neat post, Ziggy. The histograms do all show a strong similarity.

Its all in knowing and understanding your tools, isn't it?

SloYerRoll

I've read the same thing about the D50 metering the histogram green chanel. On top of that, a histogram on any Nikon is not created from a raw image if your shooting RAW. It creates a jpeg to reference for the histogram and takes the readings off the jpeg and "discards" the jpeg. (even if your shooting RAW only, it still makes a temp jpeg) I'm not sure of the process, but I do know at least Nikons do histograms that way. Even when you get into the D2x. When you get into the high end Nikons, they reference all three channels but it's still referencing a jpeg. (Anyone know if Canons do the same?)

Don't get me wrong. I'm a huge advocate of the histogram (I'm glued to it in post). But the hardware I use (D50) in regards to the histogram is inferior to a point.

@Rustic: That's why I dial everything in and thenuse the highlights (or blinkies). This way I can glance at the LCD and know if I blew anything out. I rarely want to sit there and get really picky on the histogram. This may indeed change after I get anicer body, but for now, blinkies it is..

-Jon

wxwax

Good one Ziggy! It shows what I've often blundered into: the camera shows there's room for more light, but the computer software disagrees and tells me I've lost some highlights.

The computer always wins, so when I'm worried about highlight detail, I leave quite a bit of room to the right of my camera's histogram.

i_worship_the_King

What does it mean when the 'peaks' of the histogram seem to run off the screen? (Basically, what should have been labeled on the axes of these graphs?) In addition, what are some common errors to look for. I'm aware of everything bunched to the left underexposure and everything to the right overexposure - what about the heights of the peaks?

schmoo

If the peaks go past the top of the histogram, you're losing details in those areas. Folks use the term "clipped" or "blown."

i_worship_the_King

I know this is asking a lot, but does anyone have an example of this?

I'm getting pretty competent at overexposing and underexposing - but are you saying that you can overexpose when the histogram is all in the middle of the screen? Or is it something like you're losing a shade of a color.

rwells

i_worship_the_King wrote:

...are you saying that you can overexpose when the histogram is all in the middle of the screen? Or is it something like you're losing a shade of a color.

No and No


From left to right the histogram shows luminance from complete black (left side) to complete white (right side)

If there is a gap on both ends, then you are not loosing any details no matter how high the peaks are. The peaks are simply showing how many pixels are luminated in that particular range.

Now didn't that sound confusing???

OK, visualize this: shoot an 18% gray card. What the histogram will show is a very skinny line from bottom to top of the histogram and it should be in the middle if exposed correctly. (this assumes a very evenly lit gray card - otherwise you will have a bellcurve due to the card having a hotspot, then the light tapering off from that)

The skinny line from bottom to top simply tells you that ALL of the reflectance (luminance) in the shot is within that narrow range.

Clear as mud?

Now, go back and read all about the histogram from the links posted. Believe me, it will sink in.

i_worship_the_King

^hahaha, already read them and more. What you said in the first few lines sank instantly in. What they need is LABELS on the graphs

So what I'm looking at is Amount of pixels vs. Luminence.

Or something like that... FINALLY I CAN STOP WORRYING!

Thanks for the help everyone!

Phil_Flash

RGB Histogram
Newbie's First Post: (please excuse my ignorance!)

I shoot (in Av mode using "Neutral" Picture Style) using the Canon XTi's RGB histogram. Normally, I try to "expose to the right" per the LL site, but I have noticed that the green channel is usually clipped in the shadows. This happens even if I try to center the histogram using Av adjustments.

If I adjust exposure to try to shift the green channel toward the highlights, the R and B channels clip in the highlights. What gives?

I can work with the G channel in DPP and adjust, but then color balancing become a bit arduous.

For you XTi Experts: Should I try to use the White Balance Correction matrix adjustment to handle this?

Thanks in advance,

Phil

rwells

Phil_Flash wrote:

Newbie's First Post: (please excuse my ignorance!)

I shoot (in Av mode) using the Canon XTi's RGB histogram. Normally, I try to "expose to the right" per the LL site, but I have noticed that the green channel is usually clipped in the shadows. This happens even if I try to center the histogram using Av adjustments.

If I adjust exposure to try to shift the green channel toward the highlights, the R and B channels clip in the highlights. What gives?

I can work with the G channel in DPP and adjust, but then color balancing become a bit arduous.

For you XTi Experts: Should I try to use the White Balance Correction matrix adjustment to handle this?

Thanks in advance,

Phil

Hi Phil, welcome to Dgrin


If your clipping on the left side (any channel) and you can't push it to the right side without clipping there, your scene simply has more latitude than your sensor can handle.

There are some band-aid things to try, but for the most part under normal "street" shooting conditions you just have to decide what is more important in the scene and let the less important side clip. (shadows or highlights)

I don't want to get into the whole (additional/control lighting thing here - not on-topic for this thread)

One thing that helps sometimes is to lower the contrast and saturation settings in-camera. Test this out sometime.

Take a scene that easily fits within the histogram, decent amount of room on either end. Take this shot with your normal settings. Now take the same shot adjusting the contrast, then the saturation and see for yourself what happens on the histogram. (both up and down) This is a good way to visualize how much or little gets changed in the "width" of the histogram data.

You can bump these things later in post production, but you can't add back in data (clipped) that isn't there.

I don't know if the XTi has "Picture Styles", but if so, you can make a custom style that has the contrast/saturation lowered. Then, as needed you can quickly switch to that style.

If its a shot that you can use a Grad ND filter, that will lower the scene latitude, but generally were talking about landscape pics here.


There's lots of great photogs here that will probably jump in to give you some more info. (and probably more useful)

Phil_Flash

rwells wrote:

{snip...One thing that helps sometimes is to lower the contrast and saturation settings in-camera. Test this out sometime.

Take a scene that easily fits within the histogram, decent amount of room on either end. Take this shot with your normal settings. Now take the same shot adjusting the contrast, then the saturation and see for yourself what happens on the histogram. (both up and down) This is a good way to visualize how much or little gets changed in the "width" of the histogram data...
{snip}
...I don't know if the XTi has "Picture Styles", but if so, you can make a custom style that has the contrast/saturation lowered...{snip}

Thanks, Randy!

Using DPP and a RAW (.CR2) image, I can see that lower contrast will "tighten-up" the RGB histogram. In addition, I also saw how lowering saturation brought the R, G, and B profiles more in line with eachother.

The XTi does, indeed, use Picture Styles; I have been shooting in "Neutral" which, on the XTi, has Sharpening set to zero, and Contrast, Saturation, and Color Tone set to their midpoints. I much prefer to tweak these in DPP after the fact.

So, I will tripod-mount my camera, pick a scene which models the blown G channel (but with the R and B channels w/i the RGB histogram boundaries) and see if I can bring the all three channels within the boundaries using a combination of decremental values for Contrast and Saturation (custom) Picture Style adjustments.

Do I have this right?

Thanks,

PF

rwells

Phil_Flash wrote:

Thanks, Randy!

Using DPP and a RAW (.CR2) image, I can see that lower contrast will "tighten-up" the RGB histogram. In addition, I also saw how lowering saturation brought the R, G, and B profiles more in line with eachother.

The XTi does, indeed, use Picture Styles; I have been shooting in "Neutral" which, on the XTi, has Sharpening set to zero, and Contrast, Saturation, and Color Tone set to their midpoints. I much prefer to tweak these in DPP after the fact.

So, I will tripod-mount my camera, pick a scene which models the blown G channel (but with the R and B channels w/i the RGB histogram boundaries) and see if I can bring the all three channels within the boundaries using a combination of decremental values for Contrast and Saturation (custom) Picture Style adjustments.

Do I have this right?

Thanks,

PF

pathfinder

Phil,

I am curious - you say the green channel is under exposed badly, and the red and blue are overexposed. What exactly is your subject? What is its lighting? Or, are you saying that this always happens no matter what you are shooting?

And how does your final image look, or are we studying only histograms here?

It is good to remember, that the Bayer algorithm is set up so that 60% of the pixels in our cameras are green pixels, 30 % are red, and only 10% are blue. As Randy said, frequently out of doors in natural lighting, the scene contrast range may far exceed the contrast ratios our sensors are capable of recording. You either lose some shadow detail ( deep dark shadows) or lose some highlights. Think of a forest with the sky and clouds visible. Very bright highlights in the sky, but very dark, indeed, down under the leaves at the base of the forest.

I would not think that a neutral subject should routinely under expose the green channel and over expose the red and blue channels. That is not my typical experience. In the end, we want fine images, not fine histograms, don't we?

Addendum: Ziggy has been good enough to correct my error about the Bayer Algorithm. The pixels in Bayer type sensors are not 60% green, 30% red and 10% blue ( I think those ratios are the way Photoshop converts a color image to a gray scale image)

As Ziggy pointed out to me in a PM, the correct ratio is 50% green, 25% red, and 25% blue.

Thanks,Ziggy, for keeping us on our toes and the straight and narrow!!

Phil_Flash

pathfinder wrote:

...I am curious - you say the green channel is under exposed badly, and the red and blue are overexposed. What exactly is your subject? What is its lighting? Or, are you saying that this always happens no matter what you are shooting? ... And how does your final image look, or are we studying only histograms here?

Let me be a little more exact. In a scene that includes a bright (overcast) sky above a forest of trees in New England leaf, with a pond in the foreground, my R and B channels are within the boundaries. However, if I look at the G channel, it is blown in the shadows.

The image looks good after processing the RAW file in DPP, but I am just trying to eek-out what's blown in the green channel...perhaps to no appreciable difference in the final image, I grant you.

Also, I understand that by shooting to the right, I am utilizing more of the tonal range of my sensor. So (theoretically) my thoughts are that if I can get more of the channels to the right without blowing the highlights, I might see some image difference. Thus, underexposing in RAW will shift channels right but, in effect, I am just trying to shift the green channel a bit.

However, a predominance of green sensors certainly can account for this in a scene with a wide range range of light.

In the end, it probably does not make much difference as the resulting image still looks very nice!

pathfinder wrote:

I would not think that a neutral subject should routinely under expose the green channel and over expose the red and blue channels....

Agreed. This issue does not occur with the green channel when the scene is more neutral in nature.

Thanks!

PF

pathfinder

Phil_Flash wrote:

Let me be a little more exact. In a scene that includes a bright (overcast) sky above a forest of trees in New England leaf, with a pond in the foreground, my R and B channels are within the boundaries. However, if I look at the G channel, it is blown in the shadows.

The image looks good after processing the RAW file in DPP, but I am just trying to eek-out what's blown in the green channel...perhaps to no appreciable difference in the final image, I grant you.

The reason the green channel has areas of under exposure in the scene you described, is that if the exposure is going to be correct for the blue sky, the dark shadows at the base of the forest are very dark, and what little light they get is reflected from the nearby leaves - hence the green channel is underexposed ( blown) on the left side of the histogram.

Also, I understand that by shooting to the right, I am utilizing more of the tonal range of my sensor. So (theoretically) my thoughts are that if I can get more of the channels to the right without blowing the highlights, I might see some image difference. Thus, underexposing in RAW will shift channels right but, in effect, I am just trying to shift the green channel a bit.

However, a predominance of green sensors certainly can account for this in a scene with a wide range range of light.

When you "shoot to the right" in the histogram, you are maximizing the exposure as much as possible without blowing the highlights. This is not under exposure, but increasing the exposure slightly or planned "over exposure".

When processing such a file shot 'to the right', you may need to slide the Exposure slider to the left in the RAW converer to bring the brighter areas back to a more appropriate level of brightness in the final image. This may help decrease the noise in the image slightly.

In the end, it probably does not make much difference as the resulting image still looks very nice!

That was what I was suggesting when I asked if we were discussing a specific image and its histogram, or just discussing the fine points of a histogram of an image which never gets printed to paper.

ALL images should have some small areas of under exposure - we call those area of severe under exposure "The Black Point" and setting a Black Point Limit is one of the steps of post processing an image, along with setting a WHite Point as well. The WHite Point defines what parts of the image will receive little to no ink on the paper when printed eg: white point.

Phil_Flash

Very informative, pathfinder. Thank you so much for helping me to figure this all out!



PF