I hope this helps:
If you are modeling pipelines using the UML 2.x (presumably in EA), components, ports, connectors and composite structure diagrams are your best friends. Perhaps you could model it as a digital signal processor?
You could create an archetypal component named a "processing unit" and create a number of specialised (child) processing units that derive from this component. Each unit would be for one image filtering function that your pipeline would use (perhaps a sharpener, or even a colour channel spiltter).
You could then create an archetypal class that is specifically for typing ports on your new components.
Next, you should create specialised ports of your archetypal port class. Your specialised ports are for supporting the different kinds of imaging information that your pipeline would transfer to and from components.
Next, create ports on the specialised processing units you made earlier. The next step is to compose your processing units in a "pipeline" component and connect them together.
You then need to create a simple "pipeline" component, which will compose all the components. This could be another processing unit instead, if you want to support composition of processing units.
Add to this pipeline (using "add parts..." from the component's context menu) the processing units that you require.
Then, on each processing unit that you've added, you need to expose its ports (by showing it's "embedded elements" from its context menu).
Finally, you need to model the connections between each processing unit's port with connectors.
The pipeline, or other type of component that composes all of your processing units, could expose ports as well.
In this case, you should using connection "Delegate" connectors which connect from the outer component's port to the child component's port.
Of course, the SysML (
http://www.omgsysml.org/) is a much more suitable language for your purposes.
