Our histogram represent the distribution of tones and, since it shows the distribution of the camera's generated jpeg, we have tonal distribution from zero to 255 (an 8 bit image) - detailess black to detailess white.
We also know that digital information is directly proportional to brightness so, if we choose a dynamic range of five stops, the brightest stop will have one-half the information (128 to 255), the next lower stop one-half of that (64-127), etc etc. So, we push the histogram to the right (avoiding clipping) to capture all the information (or as much as possible)
Here is the confusion:
Say we're in aperture priority at f/8. The dynamic range is then from f/4 to f/16 - five stops centered on f/8. The camera chooses a shutter speed such that it reads medium tonality at f/8. But, if the 128 to 255 range is at the brightest stop, 128 is not captured at f/8. So, do we come to the conclusion that metering and our histograms are not related?
To make things even more difficult, most discussion about information capture vs stops of light show the histogram superimposed on a linear (five equal spaces) of stops when it should probably be a logarithmic distribution with the brightest stop taking up one half of the display. Then when you superimpose a straight line on top of this to show information distribution, the mid-point of that straight line will intersect the mid-point of the histogram and every thing looks right. An alternative would be to show our five-stop range in equal spaces but to show the information curve as an exponential distribution. Again, the mid-points of the information curve and histogram will intersect.
Okay - okay, we're photographers not mathematicians but has anyone else been thinking about these aparent discprepancies?
George
Full Version: Histograms
QUOTE(gtheodore @ Nov 13 2006, 05:23 PM)
Our histogram represent the distribution of tones and, since it shows the distribution of the camera's generated jpeg, we have tonal distribution from zero to 255 (an 8 bit image) - detailess black to detailess white.
We also know that digital information is directly proportional to brightness so, if we choose a dynamic range of five stops, the brightest stop will have one-half the information (128 to 255), the next lower stop one-half of that (64-127), etc etc. So, we push the histogram to the right (avoiding clipping) to capture all the information (or as much as possible)
Here is the confusion:
Say we're in aperture priority at f/8. The dynamic range is then from f/4 to f/16 - five stops centered on f/8. The camera chooses a shutter speed such that it reads medium tonality at f/8. But, if the 128 to 255 range is at the brightest stop, 128 is not captured at f/8. So, do we come to the conclusion that metering and our histograms are not related?
We also know that digital information is directly proportional to brightness so, if we choose a dynamic range of five stops, the brightest stop will have one-half the information (128 to 255), the next lower stop one-half of that (64-127), etc etc. So, we push the histogram to the right (avoiding clipping) to capture all the information (or as much as possible)
Here is the confusion:
Say we're in aperture priority at f/8. The dynamic range is then from f/4 to f/16 - five stops centered on f/8. The camera chooses a shutter speed such that it reads medium tonality at f/8. But, if the 128 to 255 range is at the brightest stop, 128 is not captured at f/8. So, do we come to the conclusion that metering and our histograms are not related?
This must be a FAQ by now.
The brightness values you cite are before gamma has been applied, also known as "linear mode" when you preview your raw conversion.
The histogram is based on the JPEG after gamma has been applied.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.