In between structure and texture: Strexture

If we have a set of images of different examples of the same basic object (human brains, human faces, cars etc), then there are already ways to automatically locate the common structure across the images. So for faces, the basic structure of two eyes with a nose in between, mouth below the nose, and chin below that, is common to all normal human faces. This commonality of structure can be detected, and then used to investigate the way that the exact shape of the face changes between individuals (identity), and how it changes for a single individual (expression). Similar investigations can be applied to complicated biological structures such as brains, and can help elucidate the difference between populations (normals compared to schizophrenics, say), as well as the way degenerative diseases such as dementia alter the brain structure.

However, if we consider a different biological structure, for example beech trees, we see that although there are patterns of branching common to all beech trees (and different to, say, oak trees), this does not mean that the exact pattern of branching is the same for any two beech trees. We hence have a class of variation/similarity which is halfway between the strictly invariant structure considered above, and the more general variation that can be considered as texture. This has been given the name of strexture.

The aim of this project is to find ways of describing and modelling strexture, and its variation across populations. This is important in various biomedical applications. One exemplar which will be used during this project is the pictures of the vasculature in the human eye (retinograms). This has some regularity between individuals, somewhat like beech trees compared to oaks. It is an important test case, since retinograms are used in screening for diabetic retinopathy, the leading cause of irreversible blindness in the western world.