Image Processing and Computer Vision are very broad concepts, applicable to any number of contexts involving data acquisition from a camera of some sort. Most images studied in the context of denoising or compression are every-day, human-world images, photos of people, buildings, scenery etc. The human world has a complex structure, full of straight lines and edges, and is difficult to model, therefore many of the methods developed for such images are heuristic in nature.
In contrast, scientific imaging refers to working on two- or three-dimensional imagery taken for a scientific purpose, in most cases acquired either through a microscope or remotely-sensed images taken at a distance. In contrast to the complex structure of the human (meter) scale, at both the micron scale (crystals, grains of sand, groups of cells) and at the kilometer scale (forests, deserts) the natural world has a much more random, textured, or irregular structure which can often well be modeled mathematically as a random field.
The methods developed for scientific imaging tend to be less heuristic, and more model based, than those in photographic image processing. Markov random fields, Gibbs random fields, hidden Markov models, and wavelet models are some of the approaches which we use here.