Photography Basics: ISO and noise

714px-matrixw
CMOS sensor, by Filya1 (Wikipedia).

In previous posts we saw one of the three pillars that sustain the exposure in photography: the aperture, in a theoretical and practical way. We saw that using a wide aperture let us use more light at the cost of decreasing the depth of field, and the artistic implications of this trade off.

Now we are going to study the second pillar of the exposition: the sensitivity of the sensor. The sensitivity, usually called ISO in the cameras, represents how receptive is the pixel in the sensor to an amount of light that reaches it. It is represented by a number, usually starting on 100, but sometimes in some cameras it can start even lower: 50. The higher the sensitivity the less light we will need to achieve a concrete exposure value. This means that by increasing the ISO value we will be able to take pictures in darker conditions.

The relationship between the ISO and the light is linear, instead of quadratic, as it happened with the aperture. This means that doubling the sensitivity will effectively double the amount of light that we receive.

Example: We want to take a picture inside a church, and we correctly expose it using a shutter speed of 1/10 s, f/5.6 and ISO 100. We want to keep aperture constant because it is in the sweet spot of our lens. The problem is that the people are slowly moving inside the building, appearing diffuse in the final picture, and we want to freeze them. If we increase our ISO to 800, we increase the light 8 times (800/100 = 8), which means we have doubled the amount of light three times (Log² 8 = 3, or what means the same: we have doubled 100 three times to reach to 800: 100 –x2-> 200 –x2-> 400 –x2-> 800). We can decrease the shutter speed by 8 times to compensate, up to 1/80 s (1/10 –x1/2-> 1/20 –x1/2-> 1/40 –x1/2-> 1/80). This speed is more than enough to freeze the slow movement of the people inside the church.

Some cameras change ISO in full steps. This means that increasing or decreasing the value means doubling or halving the amount of light. In a camera configured to change aperture or shutter speeds in thirds, this means that every time we change ISO one step, we can move 3 “clicks” the other values (if you don’t understand “click” nomenclature, refer to the theoretical explanation on aperture.

Example: We have the parameters of the last exercise: 1/80 s, f/5.6, ISO 800; reducing ISO to 400 means halving the amount of light (one stop). To compensate, we should move one stop (three clicks) the aperture or shutter speed opening or slowing it: 1/40 s or f/4 (but not both of them at the same time). Also, a combination of both can be used, using three “clicks” amongst both, for example: 1/60 and f/4.5 or 1/50 and f/5. In the first case we move two thirds the aperture and one third the shutter speed, on the other case just the opposite.

Other cameras, on the contrary, also accept thirds or halves of ISO values. In these cases the same rules as in aperture are valid: one click on any direction can be compensated with one click in the opposite direction in any of the other two parameters.

The ISO, seen this way, seems like a silver bullet against the lack-of-light-monster. But as always happens in photography there is a trade, and in this case is nor artistic but technical: Increasing the ISO also increases the noise in the picture. The noise is a small variance in the brightness and color of every pixel in the picture that cannot be controlled.

When a sensor captures a photon it generates a signal by releasing an electron to the internal circuit. The more electrons in the circuit, the more light the camera understands it has been captured. Sometimes, when no photon has been detected, a “rogue” electron can escape from the sensor (and goes living la vida loca). The camera doesn’t have any way to know that this one wasn’t released by light, so it understands it as that. When we increase the ISO value the sensor emits two, four, eight… electrons for each photon, but also when electrons escape by themselves they do in that same amount. This means that in a black pixel, when no light is received, the theoretical color should be black (0% grey), but it usually has a very small value of grey (e.g. 0.5% grey), because of the electrons that escaped without light interaction. When we increase the ISO, the pixel still doesn’t get any photon, but instead of releasing 1 electron, releases 8 each time, so we obtain 8 times more grey than before (a 4% grey).

The release of these electrons is completely random. If it wasn’t we could simply adjust brightness, contrast or levels in the final picture to compensate. But we cannot really know which pixels have noise, or in which amount, as it varies all the time. Instead of having a “mist” covering the picture we get a random “snow” that changes from picture to picture. Also, every white pixel is formed by 4 basic ones (1 red, 2 green and 1 blue), and not all of them emit the electrons at the same time, which means that every pixel of noise has a different color. As a result, increasing the ISO decreases the quality of the picture.

Also, from the previous explanation we can deduce another characteristic of noise: it affects more to the dark areas of the picture than to the highlights. As we can see, for an existing ISO value all pixels have a similar chance to produce noise (a final value in luminosity ranging from 0 –no “rogue” electrons- to x –being x an unknown value of luminosity-). In any dark pixel, with luminosity close to 0, the action of adding that amount x increases the luminosity in a huge amount relatively to the actual value (if the real luminosity is 1% and x equals 8%, we have increased the luminosity of that pixel by 8 times). On the other hand, if a pixel in the highlights has a value of 90 %, increasing it with an 8% means an increased luminosity of 1.08 times, a very less noticeable amount compared to the surroundings.

general_X

On this picture we can see Mr. Leprechaunius Smith, Mr. Brainsqueezer and Mr. Someweirdelvenname discuss around a table about the economic theory of “hydraulic macroeconomy”. Mr. Someweirdelvenname seems to be taking the argument too seriously. The picture is taken at ISO 100 (the lowest available) with the APS-C sensor of a Canon 500D. All noise reduction characteristics are disabled for this example.

combi

This is a cutout at 100% magnification of the face of Mr  Someweirdelvenname. On the left side we see the picture taken at ISO 100. Although some noise can be appreciated on the background, in general quality is acceptable. On the right side we see the same cutout of the picture, but taken at ISO 3600. Two things are clear: The increase in noise is enormous, reducing the overall quality of the picture, and the noise is much more visible in the darker areas than in the white face. This second picture can be acceptable for printing at low size, but not for keeping cutting just a small part of it as a whole image.

 

Having the first part of the explanation in mind, some photographers who value quality tend to think on the ISO value as evil, and never increase it to avoid losing quality. This is a useful rule in general, but has a problem: Sometimes, not increasing the ISO makes the picture underexposed and darker that it should be, and afterwards it has to be correctly exposed on Photoshop. This means that all the noise that was in the darker areas is also there in the midtones once you have corrected it, as you increase the good and the bad parts at the same time. On the other side, if increasing the ISO value leads to a correctly exposed picture, the midtones might have more noise, but it will be less noticeable except in the darker areas, which means a general increase in quality. As a rule, ISO should be increased in all the cases when not increasing it would lead to an underexposed picture, and not increased in all the cases where you can achieve a correct exposure with the rest of the parameters.

Also, as an even more general rule:

Taking a low quality photograph is better than not taking it at all.

Now is time to go out and practice with the sensitivity of the sensor. Try to take some pictures on a dark room or at night, change the values and see the effects. Any questions? Want to share your experience with the ISO? Comments are open for you to contribute with anything you have to say.

Advertisements

3 thoughts on “Photography Basics: ISO and noise

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s