Ever since I first saw this little graph on my first digital camera it has fascinated me. I have always used it as an indication of the level of clipping that has occurred but I am sure it has more uses than that so this chapter should be interesting.
Firstly… what is the histogram?
Basically it is a representation, in graph form, of the light value that each pixel on the sensor sees. Here I have put on my kitchen table the inside cover of our course note folder which is grey with a bit of black mounting board on one side and some white paper on the other. Below it I put the histogram that this picture produces in CS4.
On the histogram the horizontal axis represents the level of brightness, black on the left and white on the right. The vertical axis represents the number of pixels. The histogram shown here is common to most that I have seen in that it restricts the top of the graph so with a lot of a single brightness level the spike will be cut short by the limitations of the graph display. So, whereas it might be thought that there is a lot of grey and white but only a little black we know from the photo that this is not true. What this histogram is showing us is that the black is very pure black so the spike is narrow whereas the grey and white are spread over several shades which is why the spikes are broader. Were we able to see the top of the graph we would find that the black spike is taller than either the grey or white since the pixels seeing those shades are spread over a wider area.
That aside, we can tell that both the black and the white are within the dynamic range of the camera since the spikes fall within the edges of the graph. Should either spike go beyond the sides of the graph we would know that some clipping has occurred, in that the camera hasn’t been able to record light levels that fall outside the graph. Those areas where this occurs will look plain white or black.
The exercise asks us to shoot the three most basic categories of scene by contrast showing the histogram and an example of the changes that occur with +1 and -1 stop of Ev.
Low Contrast Ev0:
This scene is dominated by the plastic ball the girl is looking through. Apart from the hole the plastic has given the rest of the image a very similar light value so the histogram has a definite peak which shows this. The darker area through the hole gives rise to the smaller bump to the left of the histogram.
At +1 stop the histogram has moved to the right (the bright side) and now falls off the side of the graph showing that there is some clipping of the dynamic range in the brightest areas of the picture. The small bump representing the dark area has flattened a little as the change in light has brought more detail out of that area and it now covers a greater dynamic range.
At -1 stop the main hump of the histogram has predictably moved left and the smaller hump has grown a little and shows some clipping at the darker end of the dynamic range.
Moderate Contrast, Ev0:
This shot of the Empire State Building reflected in an office block has moderate contrast. This is seen by a fairly flat histogram that extends the length of the graph and only spills off the ends by a small amount. The vast majority of the photograph falls within the dynamic range of the camera and is covered by the histogram.
At +1 stop the graph has moved to the right and now we can see that there is a peak on the right edge of the histogram which indicates clipping of the bright areas of the image.
At -1 stop the graph has moved left and the clipping of the bright areas has been cured but now we have the same problem at the dark end of the dynamic range. This is a clear graphical demonstration of the need for accurate exposure for subjects which cover a large dynamic range.
High Contrast Ev0:
For High Contrast shots the histogram forms a U shape with peaks at either end of the scale showing that there are a lot of extreme darks and lights within the shot. I have expanded the size of the graph for this set for clarity. So the dark hair and T shirt of this hi-key portrait are all in the very low end of the range whilst the hi-key background, being bright white, is at the very high end of the range.
At +1 stop the histogram has moved right as we expected and from the graph it looks almost as if there is no bright whites at all! We know that this is untrue just by glancing at the photo. It looks this way because most of the white has been pushed off the end of the scale. However, when I took the screen shot I rested my cursor on the right edge of the histogram to get a count of the pixels at position 255 (the far right). This indicates that there are 17,975,601 pixels at this level, an enormous number and easy to miss because the peak is so narrow and right at the edge of the graph.
When we look at the -1 stop image it is obvious from the photo that the bright white background is showing a bit of fall off in the bottom right corner. This is backed up by the number of spikes that are showing at the right end of the histogram. The count at position 255 is still very high but 7 million less than +1 stop image. As we expected the graph has moved left and the mid tones have taken up a stronger and darker presence.
The histogram is a valuable tool when shooting in difficult light conditions but it has limitations. The most serious one is the lack of impact that the graph shows for a highly over (or under) exposed image as it only appears as a thin line at the far end of the scale. As can be seen in the Ev+1 high contrast portrait above, the vast amount of over exposure is not obvious from the histogram. A very useful tool to back up the histogram on the camera back is the highlight alert. This setting causes the areas of highlight clipping to stand out by being outlined or in the case of my camera by flashing. Now the photographer has a visual clue to back up the histogram as to where and how much of the image is beyond the dynamic range of the sensor.