The first part of this exercise was to take similar shots at both normal and high sensitivity, in a situation that was marginal – i.e. where the mixture of light level and subject movement was only just possible.
All images were taken on a Canon 5D with a 24 – 105mm lens. A tripod was not used due to being too intrusive for most of the locations. Images have been taken in RAW format and reduced in resolution for web purposes.
The first two shots below, due to size and low resolution, look identical which slightly defeats the object of the exercise. I’m viewing them on a 24 inch screen at 100% magnification and can see differences. At ISO 100 details are slightly soft, and shadowy areas lack the colour saturation of ISO 640 image. Colours are also more true to the originals in the scene.
I usually shoot with fairly low ISOs, so increasing it to 640 was a walk on the wild side.
The images below are close-ups of two areas from the above images.
Again it’s only when the image is viewed at a relatively large size that visible differences become apparent. Finer details can be seen in the higher ISO image, and colours are slightly brighter.
Shaded areas also produce greater clarity.
Not convinced so far that the higher ISO has made a staggering amount of difference, I increased it to 800. I still had to enlarge the RAW image to 100% before any real visible differences could be seen. The amount of grain was quite apparent at the high ISO, and dulled highlights as a result.
The shaded areas of the image suffered at the higher ISO and lighter areas show the typical speckled effect that noise produces. This was particularly evident in paler tones which are effectively covered in a grain several shades darker than the actual tone.
Enlarging the following images revealed darker skin tones at the higher ISO due once again to noise levels, and paler colours appeared muddy for the same reason.
Also enlarged, the next images have experienced loss of quality, especially in highlighted areas as before. Detail is more sharply defined at the lower ISO range, and noise levels at the higher end of the scale have the effect of darkening colours compared to the original tones.
Below at a small scale, there is relatively little difference between the two images apart from slightly lighter tones again in the paler colours.
Even enlarged, the differences are barely perceptible. Slightly less detail becomes apparent with the higher ISO in lighter tones, and conversely greater contrast between objects and structure can be seen in darker areas.
I decided to try the same experiment in brighter light, and increased the higher ISO to 1600, convinced that this would show a marked difference in noise. As seen below, with the smaller size you would have to look very closely. Results show very marginal colour, and contrast differences.
Enlarging one area does however show a greater difference in image quality with the higher ISO. The effects of noise are similar to pixelation when an image is enlarged beyond its optimal capacity.
The overriding result is that smaller sized images at a high ISO can appear ‘almost’ identical to the ones at the lower end.
For most people who print no bigger than an A4 size, this may be a happy compromise when higher ISO settings are required due to reduced light levels.
Colour does seem to change from the original tone as ISO numbers rise, and highlighted areas darken and appear muddy, which may not be an issue for some subjects. Skin tones for example may not stand up to such high settings.
I can see that reflective surfaces may also suffer considerably as highlights are dimmed due to noise levels increasing, again effecting such things as catch-lights in eyes if portraits or people are the main subject.
For some images noise levels may not be an issue, and depending on choice of paper when printing, may even be an advantage.
The general evidence though seems to support the theory that higher ISOs and the resulting noise does reduce quality and accurate colour rendition.