I've been using my 10D for several months now and have been very happy with it in controled situations. However, with wedding season upon me I've been testing it with more variables (in regard to lighting) and now I'm feeling less confident. I realize the LCD is not accurate for exposure, so I move to the histogram. However, I really don't feel confident with the histogram either. I understand that the histogram is a visual representation of a given exposure, but I don't know how to judge if if the exposure is accurate or not. Obviously the overexposure warning is very helpful but, if I'm photographing a black cat in a coal mine the histogram will obviously look different than if I'm photographing a snow ball. Can anyone give me some pointers on the histogram or is this all by feel? I have looked at several histograms from accurate exposures and I'm failing to see the common thread.
White Balance so easy, even our 5 year old can do it.- Melissa Strickland
A histogram is not that difficult to read, but can take a little practice to interpret the results. The 10D does not break down by color, so what you are seeing is a representation of greyscal values from pure black to pure white (left to right). For RAW images this covers 12 bits (4096 values), for JPGs 8 bits (256 values).
The height of the curve indicates the relative quantity of that value within the frame. So, a pure white image would have a sinlge solid line on the far right, a pure black would have that line of the far left, and pure middle grey a line in the middle. In a wedding shot (or other high-contrast scene), assuming white gown and black tux, you would likely see a curve like a cursive letter U. High on either end dipping in the middle (depending on the background scene). Now this is an estimate on my part, and will vary on exposure and other considerations.
Take some test shots in a controlled environment, maybe using some Barbie/Ken dolls in formal wear. Or a rectangular piece of white and black paper in front of a larger grey sheet (the grey to represent the 18% grey average for a normal backdrop). Do some test exposures and see how the histogram looks for a properly exposed scene similar to that which you will expect.
I would generally suggest using an ambient light meter for high-contrast scenes, or find something comparable to middle grey to meter off of using the partial meter mode. If using E-TTL flashes, just bracket as normal. Of course, shooting a wedding I assume this is unecessary info for you and do not mean to question your experience. Just following my train of thought to the end...
Michael Feinberg
White Balance so easy, even our 5 year old can do it.- Melissa Strickland
If your review of the histogram reveals data crammed up against the far right side, you have overexposure - dial in some negative exposure compensation and shoot again.
If the histogram has a lot of space between the rightmost data and the far right side, your image is likely somewhat underexposed and you *might* want to dial in some positive exposure compensation. You might choose to just fix up the image in "post processing" however.
Bear in mind you can fix up a good deal of underexposure in Photoshop, but data is lost in overexposure and there's no getting it back. So, err on the dark side.
-Noel
White Balance so easy, even our 5 year old can do it.- Melissa Strickland
In lay terms: I think of histogram as a hill. I don't care what shape hill is. I just don't want to chop off left side or right side of hill, or if unavoidable, I minimize chopping by reshooting & shifting hill left or right. Ideally, the right edge of the hill is just to left of right side of histogram?
Michael, Noel, & all,
Is overexposed white off the histogram or stacked on the right side?
Is pure black with no detail off the histogram or stacked on the left side?
White Balance so easy, even our 5 year old can do it.- Melissa Strickland
By definition, over/under-exposed are values outside the possible range. These end up being recorded as 0 for under, or 255 (4096 in 12 bit) when over-exposed as the system *will* record somthing and those are the most extreme that can recorded. So, those should show as extreme stacking on either end of the histogram, which indicates pure black or pure white. The assumtpion is that data is lost as rarely, if ever, do you come across either extreme in the real world.
As for the hill analogy, it only works under "standard" conditions where contrasts and exposure fall within the "norm". In extreme contrast conditions, such as you would have shooting a white dress and black tux, you are more likely to see "camel humps". The greater the separation of those humps, the higher the contrasts. It, of course, depends on the lighting and area of the image taken up by the formal outfits.
As an example, most of my 4th of July images show a strong spike on the very far left (lots of dark sky), and a long horizontal bar only a few pixels high going almost all the way to the right (varies depending on the shot). This is correct because there was a lot of black and a very little bit of color (relative to the image size) that varied in intensity.
Michael Feinberg
White Balance so easy, even our 5 year old can do it.- Melissa Strickland
Thanks everyone for the information. I'm kind of gathering that the histogram is just another broad indicator of exposure, not meant for minor tweaking(?). With more experience I think it will be more useful though. I'm able to recognize when I'm really off on exposure, but it's the 1/2 stop tweaks that I just can't really see.
Coming from a MF system I prefer to work with a hand held meter. But with the spontaneous nature of weddings, I often shoot with 35mm and use its internal metering which is not as consistent as my sekonic meter. With film, of course this was fine, but with digital I want/need more accuracy. So I'm hoping to be able to learn to see problems with the histogram.
Or if anyone has any advise on how they mantain exposure accuracy when you must rely on the camera's meter...that would be nice to hear also.
Thanks again!
White Balance so easy, even our 5 year old can do it.- Melissa Strickland