The idea is to demonstrate how multiple digital captures of a single film frame, at different offsets, and then re-registered and merged, will improve what you would otherwise get from a single capture per film frame.
Image A is a random picture of a movie star off the web. An attractive image but for the sake of the test an arbitrary image. This is a stand in for the image to fall on some grainy film.
To this image I digitally added noise: image B. This is a stand in for the resulting film image (a grainy/noisy one).
Images C and D are a re-sampling of the noisy image (B) at half the definition of B. These are stand ins for what happens during transfer of film (B) to digital (C and D)
It is important to note that both C and D are exactly the same definition as each other, ie. represent use of the same definition camera used to transfer a film frame to digital.
The first thing to note is that both C and D are less noisy than B. This is purely a function of each having re-sampled the noisy image at a lower definition than the noisy image. It is called a "low pass filtration" of the noisy image. In and of itself this is not of direct interest here. It is just what happens when you use a camera of a particular pixel definition. What is of interest here is the difference between C and D.
C is the result of sampling B just once.
D is the result of sampling B four times, at different offsets, re-registering, and then merging the results.
As can be seen, D is less noisy than C. Both are from the same source (B). Both did so at the same definition. No noise removal algorithms were used. The only difference is the number of samples that were done. The important point is that the decrease in noise, between C and D, rerpesents what can be achieved using the same definition camera on the transfer of the same film.
