Histogram question.

[KenC]

The first problem is to define what you consider "underexposed or overexposed" - I'm not being difficult here (well, not just) - even aside from deliberately low-key or high-key exposures there could be a pretty wide range of acceptable histograms. Do you want to avoid having any blown highlights, even if relatively small? Do you want to avoid having any of the shadows block up? Is the scene mostly light or dark? In the case of avoiding blown highlights, this is difficult, as discussed in another thread I saw earlier, because a small overexposed area will not be evident from the histogram or the overexposure "blinkies." For a generally light or dark scene, the histogram should be a pretty good indication of whether you've reproduced it accurately (assuming that was your intention).

 

[unpopular]

The first problem is to define what you consider "underexposed or overexposed"

I think one really bad habit people get into is that they view exposure as a sort of uniform thing - the whole of the image is over or under exposed, rather than seeing images in terms of tonal regions which might be rendered too light or too dark. For example, if you were to photograph a sunset over the water the water region will appear dark if the sunset is properly exposed, this will look fine and render accurately with plenty of detail in the clouds. However, if then you spot a pelican holding a kitten in it's beak while landing on a turtle out on a rock, but in all your excitement had neglect to release/adjust the exposure, the photo will be seen as "too dark" even though the clouds and sun rays are perfect.

Images are not uniformly 'properly exposed' the shadows will naturally be darker than the well lit areas, and if important detail exists in those shadows it may not be visible if the well lit regions are 'properly exposed'. I don't like the idea of a "proper exposure" because what is "proper" depends on what tonal region important information resides, and how that information ought to be rendered to be either accurate or intent-fitting. Saying a 'proper exposure' implies that there is some objective measure that goes beyond being 'too dark' or 'too light' which I believe is entire subjective measures.

 

[Ysarex]

First: Hoping for the best with the new baby!

Second: You're talking about the histogram displayed by the camera. Whenever you talk about the histogram displayed by the camera you have to include a critical qualification: The camera histogram does not provide direct data about the exposure of the camera sensor. This is really important. The histogram the camera displays is derived from the JPEG file processed by the camera software.

If you're shooting JPEGs and the camera JEPG is your final goal, then you can look at that histogram and consider the data displayed as valid. HOWEVER if you are shooting and saving the raw files then that histogram really isn't giving you directly accurate information. When discussing the camera displayed histogram you gotta make it clear whether you're goal is a camera JPEG or raw capture. The histogram only directly applies to the camera JPEG.

An example:



That's the camera JPEG -- SOOC. I've inset the histogram for that photo (Photoshop). The histogram displayed by the camera would have appeared much the same. Based on that histogram it's fair to say the highlights are blown and the photo is overexposed. Looking at the photo of course the highlights are blown. There was color and clouds in the sky and I'm not seeing them in this photo.

Here's the histogram of the raw file that was saved for that photo:



And in fact the highlights are not blown and the photo is not overexposed and there's actually room for a smidge more exposure. The camera will not give you a histogram for the exposure of the sensor. It will only give you a histogram for the software processed JPEG. CAVEAT: In all fairness I didn't make any attempt to modify the camera's JPEG output because I couldn't care less what it does. Had I done that by applying an image adjustment contrast reduction this example could be less extreme. But the point stands; the histogram on the camera isn't telling you directly what you've got in the raw file.

So I saw color and clouds in the sky and I photographed color and clouds in the sky. Even though the camera histogram would have told me I'd blown it (I don't bother to look), I knew I hadn't.

Joe




P.S. I know about UNIWB and don't consider that a viable option, especially since I really don't need it -- I have a light meter.

 

[amolitor]

Not that I want to oppose deep ending on some technical discussion, but I think the OP was just looking for a bit of intuition about stops of exposure versus the histogram.

 

[unpopular]

If you're shooting JPEGs and the camera JEPG is your final goal, then you can look at that histogram and consider the data displayed as valid. HOWEVER if you are shooting and saving the raw files then that histogram really isn't giving you directly accurate information. When discussing the camera displayed histogram you gotta make it clear whether you're goal is a camera JPEG or raw capture. The histogram only directly applies to the camera JPEG.

I normally just use the blinkies because I do so much raw processing in my workflow that the preview is pretty worthless, I do use the histogram to estimate how far to ETTR if I misjudge the brightest region.

I have never found that the binkies to be inaccurate. Is it possible that, at least in my a700, that the blinkies reflect raw data?

 

[Ysarex]

If you're shooting JPEGs and the camera JEPG is your final goal, then you can look at that histogram and consider the data displayed as valid. HOWEVER if you are shooting and saving the raw files then that histogram really isn't giving you directly accurate information. When discussing the camera displayed histogram you gotta make it clear whether you're goal is a camera JPEG or raw capture. The histogram only directly applies to the camera JPEG.

I normally just use the blinkies because I do so much raw processing in my workflow that the preview is pretty worthless, I do use the histogram to estimate how far to ETTR if I misjudge the brightest region.

I have never found that the binkies to be inaccurate. Is it possible that, at least in my a700, that the blinkies reflect raw data?

Interesting question and I don't know. My first reaction would be to say no and assume the blinkies are derived from the JPEG engine but given that you see them prior to taking the photo it's conceivable they're being generated from the meter input in which case yes. If the blinkies on the camera are generated directly from the meter reading then they would be a better indication of sensor exposure.

Joe

 

[unpopular]

I don't think they're being read from the meter, and I am only able to see them after the photo was taken on the preview. But they're never inaccurate, over/under exposure in the raw files is always on the region where the blinkies were blinking and I feel it's a reliable way to predict overexposure. Is it possible that it looks for one or more channel on the raw data which is at 0 or 255, and just overlay the blinkies onto the jpeg picture?

 

[Ysarex]

Not that I want to oppose deep ending on some technical discussion, but I think the OP was just looking for a bit of intuition about stops of exposure versus the histogram.

Oops. I think I made a classic internet thread error here. I wanted to start by acknowledging the upcoming baby, but when I went on and said "You're talking about the histogram....," it sounded like I was directing the comment to you personally. I was in fact thinking about everyone in the thread and using the pronoun "you" as in all of you. I needed a line or something there between my first comment and second comment.

Joe

 

[amolitor]

Oh don't worry! I didn't take it as personally directed at me, I was just addressing the general tenor of the thread. It's kind of wandered off But not in a BAD direction, per se!

Thank you for taking the time to correct yourself, though. It's those little touches that make society so much more pleasant!

 

[Ysarex]

I don't think they're being read from the meter, and I am only able to see them after the photo was taken on the preview. But they're never inaccurate, over/under exposure in the raw files is always on the region where the blinkies were blinking and I feel it's a reliable way to predict overexposure. Is it possible that it looks for one or more channel on the raw data which is at 0 or 255, and just overlay the blinkies onto the jpeg picture?

OK, I was thinking live blinkies which one of my cameras does. So if it's after the photo is taken then it wouldn't be from the meter. So I don't know what it's doing then. But if it works that's useful. On my Canon and Samsung cameras the histograms do not provide sufficiently accurate data to tell me what the raw file really captured.

Joe

 

[unpopular]

That is interesting. I don't trust the histogram necessarily, but feel it is pretty close. Try setting your JPEG settings to have the saturation and contrast to minimum. That seems to help the histogram's accuracy.