Sensor Linear Capture

There are inherent differences between the way a sensor (digital) responds to light and the way that film does. When light hits the sensor it does so in a ‘linear’ (digital) fashion. This means the sensor reacts in the same way, at the same rate, to light from very bright to dark. As it is digital it can either register in ‘1s’ or ‘0s’. As such its dynamic range, although very good is not a patch on our eyes. Our eyes (and in a similar way film) work in a different, non-linear or analogue, manner. Our eyes have a built-in compression that lets us function in a wide range of situations without driving our sensory mechanisms into overload. We can move from dark to light without really noticing and even if we do our eyes adjusted astoundingly quickly and very well to the environment. The sensors in cameras lack this ability to compress light; they just see it in a ‘linear’ fashion (the levels correspond exactly to the number of photons captured).

This brings us to the next exercise. To demonstrate how the camera copes with linear capture to produce the image that we see on the screen. After capture, the camera performs some strong processing, giving the image we would expect to see. However, the initial captured image does not look the same way. The first image produced is dark, much darker than we see as an end result (in effect the image is initially underexposed). This is done by the camera to reduce the effects of ‘noise‘ in the darkened areas. The camera then applies a gamma correction curve to bring a ‘normal’ appearance to the image that we end up seeing on screen. What causes noise? I found a good, elementary description on from which I have extracted the following passage;

“....the image sensor heat can increase [a potential problem shooting in the tropics] enough to stimulate electrons (“Thermal Noise”). These superfluous electrons then get mixed in with the “true” photoelectrons that are the real target of our image sensor. The analog signal (which is converted to pixels by the sensor) is therefore contaminated before it even gets to that point.
In some situations the above scenario can cause each of the photosites on an image sensor to generate superfluous signals which can then contaminate the neighboring photosites. On smaller image sensors which cram more photosites into a smaller area, this effect can be magnified.
Another common cause of noise is shooting at higher ISO settings. As these settings basically magnify the light signal, they also magnify other unwanted signals such as background interference (eg. heat sources). When you are photographing an area of low light, the background signals can be strong enough to compete with the signals from the limited light of the area you are shooting.”

First I will demonstrate what the linear capture looks like before the processing takes place in-camera.

To do this I took an ordinary, untreated (apart from in-camera of course) jpg image. The histogram shows a good spread of tonal range across the board. It is shown here with the histogram and curve for reference.
original with curve

To this image I applied a curve to darken the image. This image gives a representation of what was captured as the linear image and what the image looked like before the camera’s processor got to work on it. With the curve applied the tonal range is pushed to the left.  This means that there are very few levels left to represent the shadowed areas, which of course, has implications for the effects of noise on the image. Again, the histogram and curve are included for reference.
curve for linear

Here is the side-by-side, before-and-after comparison of both images.

Next the job was to alter the curve on the now ‘linear’ image to bring it back to look as close as possible to the original image. The major change in the image when creating the curve was the reduction of shadowed areas. After this curve has been applied, even though it is quite close to the original, it is still discernibly more contrasty and also more noisy in the shadowed areas (and would also be to a certain degree in the lighter areas too but is more difficult to see). This is due to amplifying the noise in the shadows by having to apply such a strong curve to bring it up to a ‘normal’ viewing quality.
linear after curve adjustment
It is worth noting that, if you are following the course notes there has, in fact, probably been an image missed. It states in the notes that, ‘It will look like the curve shown on the last page’ (page 32). The image on the previous page only shows the ‘U’ curve used to produce the linear image, not the one used to bring it back to the ‘normal’ image. Of course the curve needs to be the inverse of what you have just done to obtain the linear image in order to end up with a simliar looking image to that of the original.


Leave a Reply

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

You are commenting using your 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