Exercise 4.1 – the problem of reflected-light metering

This exercise demonstrates a well-known issue with reflected-light metering, the problem of calibrating to medium-grey if the scene does not average out. The exercise is useful because it caused me to demonstrate the effect, rather than just read about it and accept it.

All images were taken with a 100mm f/2.8 macro lens on a Pentax K-1. Sensitivity is set to ISO3200. All images are JPEGs, imported into Lightroom and extracted as screen-grabs complete with the histogram. The panel below the histogram is irrelevant except that if confirms that no post-processing adjustments have been made.

The subject in each case is a calibration target with white, mid-grey and black panels.

The first set of images are taken in programme automatic mode.

Auto control

In the first image we see all three tones correctly and the histogram has three corresponding spikes.

In the black panel, the camera has tried harder to resolve the weave detail, which explains the broader ‘spike’ and may explain why it is displaced slightly to the left. For the mid-grey and white panels, the spikes are almost identical. Subjectively viewing the image itself shows that each panel has been rendered as the same overall shade of mid-grey.

The second set of images were taken in manual mode. I set the exposure to render the mid-grey panel as ‘correct’ with a zero meter bias. The identical aperture and shutter speed were used for the black and white panels.

The tones are rendered accurately and the histogram spikes correspond tolerably well with those of the control image at the top of this posting.


Exercise 1.1 supplementary

After completing exercise 1.1, I was intrigued about whether the histogram variations were solely related to scene variations or were caused by the camera, either as digital noise or vagaries in the light metering. I decided to find out by repeating the exercise while taking control of as many variables as possible.

These images are shot indoors at night (to eliminate daylight variations) lit by a single 500w photoflood lamp for consistent lighting (flash might vary between exposures). ISO was set manually at 100, to minimise noise. The camera was tripod-mounted and image-stabilisation was switched off.

In this first set, I used programme-automatic metering and automatic white balance. The resulting images are identical at a ‘collect-and-jitter’ level, so I have only shown one here.


histogs 1

These histograms are much more consistent than in the previous (outdoor) exercise. Collect-and-jitter showed the three main peaks shifting slightly. Closer examination shows that the moves are no more than one pixel to left or right.

In the second set, I adopted manual exposure control and a custom white balance (3000K as recommended for photofloods) The only variable remaining should be digital noise. Again, the images were identical to each other at a ‘collect-and-jitter’ level.


histogs 2

These histograms are almost identical. Collect-and-jitter shows no left-right shifts and a slight variation in overall height, possibly a vagary of the histogram-drawing routine. Viewing the image above shows slight variations at a single-pixel level.

My conclusion is that a very small part of the histogram variation seen in the original exercise is camera-generated, but the majority relates to the very small variations in that scene.

Exercise 1.1 … the same river …

A fascinating exercise showing the camera’s ability to detect very subtle changes in the scene before it. The images below were taken at 30-second intervals, using a tripod to ensure identical framing.

This was a ‘cloudy bright’ day with the sun behind a cloud for all four exposures. There was a slight breeze, so I avoided plants with moving leaves (which would be too easy to interpret) in favour of a subject that appeared static.

Technical information: all images were shot on an Olympus E-30 set to fully automatic mode. Image capture was JPEGs, medium-resolution normal-compression. The metering has taken all four images at 1/80, f/4, ISO200


image 1


image 2


image 3


image 4

All four images appear identical at first sight. I collected all four as tabs in a single Photoshop window, which allowed me to quickly switch between them and detect differences as ‘jitter’.

There is a subtle change in the light between the first and second images; shadows below the chairs become slightly darker and cooler. The twigs in the rear container shift slightly between the second and third frames (probably a breeze). A piece of twine on the rear trellis hangs directly in front of an upright in the third frame (a white section of wire disappears – this took a while to understand). A small leaf moves from right to left through the second, third and fourth frames.

I obtained histograms by importing all four images into Camera Raw and taking screenshots. Some surprisingly large changes can be detected by my ‘collect-and-jitter’ technique.

I have stacked the histograms together (in the same order as the images) for ease of comparison.


The changes most noticeable relate to the shape of the blue peak and the details in the colour fringes on both ‘shoulders’ of the graph.

An interesting point is that the histogram changes are more obvious than the changes in the scene itself. Whether this is due to the sensitivity of the camera as a measuring instrument, or to random fluctuations of the automated metering algorithms, is unknown at present. It would be interesting to repeat the exercise using all manual settings to bypass the automation as far as possible.