[Insight-developers] Re: missing ITK functions document

[Stephen R. Aylward]

Hi,

Vikram has asked about the PhysicalObject classes, and so I thought I'd
give a brief summary to the group.

The goal is to devise a class hierarchy that can represent spheres,
cubes, connected components, segmentation results, etc in a common
framework.   In that manner, we will be able to calculate volumes,
moments, bounding boxs, convex hulls, etc; generate masks; load and save
scenes; and more...

We are proposing a class hierarchy of "PhysicalObjects." At the base
class, a specific type of representation (data container) of an object
is not specified, just the basic api.   Derived types of physical
objects will represent different types of objects and will contain data
containers.   So, a physicalObject could be an image mask (e.g., all
values between 10 and 100 in the image represents object A), a list of
points, a function, a parameterize model, or a collection of other
physicalObjects (a CSG type object).  

Jean-Philippe Guyon is working hard on this and is following the
composite-pattern design in the book "Design Patterns: Elements of
Reusable Object-Oriented Software" by Erich Gamma et al - suggested by
Luis.

A point of future discussion is what do we want at the base class level
- what characteristics should be common to every physical object.

Our first draft includes the following concepts/functions in the base
class:
- Position, orientation, and scale in physical space.
- Bounding box.
- Test if a physical point is inside a physical object.
- Resolution of the physical object (if it is a discrete representation,
how should we 	query its points)

There would be the following filters:
* PhysicalObjectToPhysicalObject filters allow objects to be combined
(CSG).
* PhysicalObjectToImage filters allow objects to be drawn in images.
* PhysicalObjectToValue filters will give you moments, etc.

We've got some other ideas on where we can go with these, but we are
going to keep it simple at first: box, sphere, tube, image-mask, and
point-list will be the objects that we hope to provide.

Comments?

Thanks,
Stephen and Jean-Philippe