Ashford University_BUS 308

The simplest data model is one that just passes its parameters as data values, so that all the variation in the observed data is actually noise. What one wants then is some kind of average that removes the noise and returns the original value.

For a nontrivial elaboration of this model, consider the problem of subpixel-resolution imaging: if you take one digital image of a scene, you're pretty much stuck with the pixels of the image itself (well, you can use Bayesian methods to get some subpixel resolution, but only if you already have a good idea what you're looking at). But if you take several images, you can imagine their pixelations as being different applications of some kind of noise to the real image, and average them to get a higher-resolution image.
Powered by