Friday, March 25, 2011

Part 2: The Histogram - Photography with Imre - Episode 34

I managed to produce the histogram episode quite quickly and overall it turned out well. I'm not going to elaborate much here other than to discuss clipping a little more, along with presenting a method you can use to help train yourself in analyzing the histogram on your camera; especially useful for those of you new to this.

A Second Take on Clipping
Now that I look back at my script, the segment on clipping could have been put more simply or at least in a different manner. First of all, if the dynamic range of a scene you are taking a picture of is greater than what the imaging sensor of the camera can handle, then the image can have clipping in the shadows, highlights or both regions. Unlike our eyes that can see a much broader range of tones, for a given exposure the camera can only capture a portion of that and this is why some dark areas of certain scenes turn out pure black on the photo and bright areas turn pure white. What is crucial to understand is that clipped areas are completely void of detail. Taking a standard three channel (red, green, blue) image file, all three channels either have 0 (zero) or 255 as their values (so R0, G0, B0 or R255, G255, B255). As long as one color channel has a different value (like R0, G1, B0), then some image information exists, which means there is a possibility to recover a bit of detail (often though, lightening very dark areas results in noise appearing and bright areas that are darkened can look somewhat unusual and washed out).

Depending on the model of camera you have, you may be able to view a histogram in real-time (e.g. live view) or after the shot has been taken during playback or review mode. Signals (light in this case) below the signal-to-noise ratio of the sensor are clipped to black (in other words, there aren't enough photons filling up the photosite(s) to register above the noise level of the electronics) and if photosites on the sensor are saturated with photons then we see blown highlights (think of a bucket overflowing with water; you might try to put more water in, but the amount of water is "clipped" to whatever the bucket can hold).

Many image editing programs also provide histograms. Clipping in most image editors refers to pixels with all color channels having either no data (all zeros) or being at full intensity (255). There's a really good definition given in the help file for Adobe Camera Raw; I love their use of "overbright" and "overdark"... gives me ideas for some horror story... "the Overdark Lord was greater than evil itself." :P

I'd also like to clarify the use of file formats in this context. For example, if you take a picture and your camera is set to save the image as both an 8 bit JPEG file and a 14 bit Raw file, the tonal range of both file formats will be identical. Now you're probably thinking that's incorrect, but think about it this way. First of all, just because the camera is using a different file format does not mean it's using a different sensor. Unless you have a funky camera, only one sensor is used to capture the image and that sensor has a certain dynamic range (usually around five or six stops). Period. So the darkest and lightest points of either a JPEG or Raw file are essentially the same (I say "essentially" because JPEG files are processed by the camera, so for various reasons, like what the contrast setting is, the tonal range might be a little different).

But why is it then that you can "pull" more image detail from a Raw file? I'll use this analogy. You have two perfectly identical loafs of bread; exactly the same size! One loaf is cut into ten slices, while the other is cut into 50 slices (and no funny business, the cuts don't waste material so your loaf doesn't change in size just because you have more slices). Now if you look at the two loafs of bread, they are still the same size --same tonal range-- but one has more detail because it was sliced into 50 pieces. Taking this back to actual images, an 8 bit file provides you with 256 (0-255) brightness levels to work with, wheres a 12 bit Raw file has 4096 levels and a 14 bit one has a whopping 16,384 levels. Because you have more data to massage, you have the ability to lighten up seemingly black areas or darken seemingly white zones, thus "pulling" out those wonderful previously hidden details. There is more great info on this if you check out the last link in the Web resources section.

Anyway, I hope that clarifies the clipping segment of the video, especially where I said, "an 8 bit per channel image can clip more easily, so-to-speak, than a 16 bit per channel image." Now you can see it means you have more data to play with in a 16 bit image versus an 8 bit one... my bad homies, didn't mean to be shiznitting with the down low.

A Little Self Training
Ok. In the video I said I'd write about a method you can use to help train yourself in understanding what the histogram on your camera is trying to tell you. On my own cameras, I've consistently noticed that if the image seemingly appears properly exposed on the camera's display, they are actually underexposed when I view them on my computer monitor. Those images appearing a bit overexposed on the cam's screen are generally exposed properly. This is why using the histogram can be helpful, but for novice shooters it might be hard to relate what a well exposed histogram looks like. In addition, since virtually every photograph has a unique histogram, it takes a little bit of experience to gain that intuition about which one looks about right. So if you think you need a little help in this area, then feel free to follow the simple steps below:

  1. Get out there and take some photos, but be methodical. First, if possible use a tripod as you'll need identical shots. Second, use the bracketing feature of your camera to take three shots varying the exposure by 0.7 (so you'll have one shot at -0.7, 0.0 and +0.7). If your camera only does half-stops, then use that and if by chance your cam does not have a bracketing feature, then adjust manually via EV compensation. At this point, ignore the histogram; all we want are three photos of a scene. Take a bunch of shots of various locations, some darker, some lighter... whatever.
  2. Download the shots into your computer, but DO NOT delete the photos off the memory card.
  3. Now open up a file on the computer and also display the same photo on your camera with the histogram feature.
  4. And you're probably getting the idea now, but keep doing the same thing for each shot, each time comparing what the image on the computer monitor looks like in comparison with the one on the camera with it's histogram.
By comparing what the photos looks like on the two devices, you'll start to get a feel for what histograms look like for well-exposed shots, along with those that didn't turn out.


As a final note I want to add that it appears I've placed a great deal of emphasis on histograms, but keep in mind that you don't always need to use them and depending on what you're shooting they might not be that useful. In addition, it's not like the exposure has to be absolutely perfect or else the photo is junk. I personally don't necessarily believe there is such a thing as a perfect exposure to begin with, as our personal tastes or creativity will alter that definition and there's a fair amount of leverage to correct some issues in editors. Nonetheless, with some time and practice, I'm sure most of you budding photogs will start to see the benefits to histograms and when they're worthy of employing.

If all goes to plan, stay tuned for the next episode on time lapse photography. Happy shooting! L8r!

Web Resources

2 comments:

  1. This post and the video in the previous post were very informative. Thank you for sharing your photography knowledge, as well as your resources.

    ReplyDelete
  2. I'm very happy to hear you found them helpful, Linda! :)

    ReplyDelete

You may comment, but if it's spam or advertising, I'll be deleting your comment and blocking your arse!