Project: White Balance and Overall Colour

The exercise as described was very similar to an exercise in the Art of Photography (described here) so I thought I would do a variation and delve a bit deeper into how my camera treats white balance.

I started with the manual. It described the camera’s approach to setting white balance with an auto setting and six presets for sunshine, cloudy, tungsten and fluorescent lighting, flash and shade. The manual gives the colour temperature corresponding to each preset. There is also the possibility to set a custom white balance or adjust by setting the colour temperature. The latter might be useful if a light source of a know colour temperature is used. The manual gives a warning about calibration differences if using a colour temperature meter.

The Fridge

The first set of pictures is of the kitchen fridge. The lighting was a mixture of cloudy daylight and tungsten halogen, with the daylight dominant. Exposure was maintained at 1/10 sec at f5.6 and ISO800 for all these. The first was with the white balance set to auto:

clip_image002

Next up was tungsten (3200K):

clip_image004

Then a custom setting was established by taking a picture of the fridge and using this to establish the white balance:

clip_image006

Viewed subjectively, the auto and custom settings have produced correct looking pictures, the tungsten version is too cold, indicating that the room lighting was not contributing much to the overall light level.

Punch

This hole punch was placed in the living room window sill, so the dominant lighting was from the cloudy outdoors. This was taken with auto, cloudy settings, then a custom white balance taken on the white woodwork:

clip_image008

clip_image010

clip_image012

The auto and custom look quite natural whereas the cloudy setting has a slight blue-green cast.

I then tried using different colour temperature settings, starting with 2500, then 6000 and finally 10000:

clip_image014

clip_image016

clip_image018

as expected the 2500 is very blue (to compensate for the redness that light of that colour temperature would have), 6000 is about right (this is the colour temperature for the cloudy preset) and 10000 is too warm.

Limitation of auto white balance

While studying the Art of Photography, I spotted that the auto setting can be fooled (see here). The next pictures explore this further.

Using the auto setting, I took the following two pictures of the blue punch and a yellow Marmite teapot lid against a green tablecloth. The exposure was the same for both pictures:

clip_image020

clip_image022

The lighting was from outside with an overcast sky so I selected the cloudy setting for these two:

clip_image024

clip_image026

This has rendered the green tablecloth much more faithfully and consistently.

The relative proportions also affect how the camera works on auto as the next set demonstrates. The first pair were taken with the auto white balance and the same exposure. The only difference is how much of the frame the green car occupies:

clip_image028

clip_image030

Then with the white balance set to cloudy:

clip_image032

clip_image034

This has shown some interesting results, mainly that the auto white balance cannot always be trusted. Not only does the lighting affect the colour of scene, the colours of the scene can affect how the camera processes it.

Probably the most accurate way of setting the camera’s white balance is to use the custom setting. I did this on some white objects in the scene for this exercise. In fact, I deliberately chose scenes with an area of white to set a custom white balance against. An alternative would be to use a grey card. The setting only needs doing once providing the lighting conditions don’t change so it can be set at the start of a shoot.

A scene dominated by bold primaries can fool the auto setting so these should trigger a warning that an alternative setting should be used.

Advertisements

Project: Dynamic Range, Scene Dynamic Range

After discovering the camera’s dynamic range in the previous exercise, the idea here was to investigate the dynamic range 0of various photographic scenes. The course notes said to take five pictures of varied lighting conditions. Here is what I took:

Cutting Down a Tree

clip_image002

This was taken in sunny conditions (as can be seen from the shadows). There was no highlight clipping in the camera display and spot readings taken showed a highlight at the saw and white parts of the tree stump at1/350 and f6.7 (EV13.9) and a shadow in the shadow of the bush at 1/20 and f6.7 (EV9.8). This was with an ISO setting of 100. This is a fairly modest dynamic range of 4.1, well within the camera’s ability.

clip_image004

Saw in Wood

clip_image006

It was cloudy by this time, so I was expecting a more restricted dynamic range. With the ISO set at 200, the highlight here was in the handle of the saw at 1/350 and f5.6 (EV12.4) and the shadow at 1/10 and f5.6 (EV7.3). A dynamic range of 5.1.

clip_image008

Stump

clip_image010

It was getting later by this time, the sun was going down and I needed an ISO of 800. The highlight on the lightest parts of the exposed stump was 1/160 at f5.6 (EV9.3). The darkest part turned out to be the area of bark to the right which recorded a spot reading of 1/15 at f5.6 (EV5.9). This results in a dynamic range of 3.4.

clip_image012

Takara

clip_image014

This picture of the family pet was taken indoors at late afternoon opposite a west-facing window. ISO was 200 and there was highlight clipping on the couch just behind her head. This gave a spot reading of 1/350 at f5.6 (EV12.4). The lowlight is the dark blanket on the couch at 1/10 at f5.6 (EV7.3). This gives a dynamic range of 5.1.

clip_image016

Takara Close-up

clip_image018

At ISO800 the clipped highlight on the left muzzle showed 1/45 at f5.6 (EV7.5) while the shadow rendered 1/8 at f5.6 (EV5.0). This gave a dynamic range of 2.5.

clip_image020

This was an interesting and useful exercise for a number of reasons. It is clear that my subjective assessment of the contrast range of a scene is not very accurate. The first image, which I considered at the time to have a high range turned out to be fairly modest. The subsequent image, which I thought was flatter, had a higher range. Also, the tonal range showed by many of the histograms do not match the measured dynamic range.

Secondly, my judgement of which part of the scene is the highlight and which is the deepest shadow in inaccurate and probably the measurement of the exposure value is not as precise as it could be.

This exercise proved very instructive and I intend to repeat it to continue to develop my eye and ability to recognise tonal ranges in a scene.

Project: Dynamic Range, Your Camera’s Dynamic Range

This exercise was to measure the camera’s dynamic range by capturing a scene with a high enough dynamic range to exceed the camera’s limits. I waited for a sunny day then did as was suggested in the course notes. A large piece of what card was placed in direct sunlight by my front door, this provided the highlight and the low end would be provided by the dark door and/or the less well lit interior of the house. The upper limit of dynamic range would be indicated by a spot metered reading on the white card and I took a number of spot metered readings of the door and interior.

A number of exposures were taken, all at f6.7 and starting with 1/90 sec shutter speed, reducing until the card showed no clipping. This occurred at 1/350 sec. This was taken as the working shot and is shown here with the spotmeter readings:

_MG_0419rl

Photoshop pixel dropper was used on the card to show that the exposure was just below the clipping point.

clip_image007

Values of R-253, G-252 and B-250 are satisfactory

The dark area was lightened and zoomed to pixel level to evaluate the noise:

clip_image009

It is apparent that it is all fairly noisy, although it is difficult to determine what is the limiting noise as far as dynamic range is concerned; there is not a lot of detail to compare it with. Nevertheless, the bottom limit can be taken as the door.

Based on this, admittedly crude, experiment, the dynamic range of the camera is about 7.1 EV. This is a bit disappointing and does not compare particularly well with the result published by DxO for the same camera (http://www.dxomark.com/index.php/Cameras/Camera-Sensor-Database/Canon/EOS-40D) of 11.3 EV.

There are a number of things to be learnt from this:

1. This is not a very accurate way of measuring the dynamic range. The highlight is fairly accurate. Although the exposyres were half a stop apart, checking the levels of the three channels showed that it was not far from full white. The lower level was more subjective. Determining the noise/detail threshold is subjective anyway, but as it happened and due to the choice of scene, there was not a lot of detail to help make the choice.

2. Despite this, the exercise was another step towards understanding how a digital camera responds to light, especially at the extremes,

3. With this understanding, it should help to evaluate a scene with a view to how the camera will capture it. This is more the subject of the next exercise.

Project: Noise

For the purposes of this exercise, I chose a scene with some detail, an (almost) textureless surface in shadow and in light and a surface in between. (more…)

Project: Highlight Clipping

With the sensor output using just a limited number of bits to represent the intensity of light falling on it, it is inevitable that there comes a point when the light is too great to be properly represented. Film and our eyes have a natural roll-off in their response so that the progress to extreme brightness is gradual and not easily noticed. Digital sensors have a linear response. This makes the transition abrupt. The purpose of this exercise is to investigate this effect by taking a series of pictures of the same scene at different exposures and comparing the area where highlight clipping occurred. (more…)

Project: Linear Response

The idea behind this exercise was to simulate the response of the sensor to light. Unlike the human eye, the sensor’s response is linear, doubling the light falling on it will double its output. The eye on the other hand is more sensitive at low light intensities, with the response rolling off as the light level increases. The exercise involved working with a single jpeg image and reversing the processing the camera applies to attempt to show what comes off the sensor. (more…)