Ok, I did some more research to see what kind of construction parts are out there. My thinking was to build a prototype using DIP parts, to make things easy, where the circuit would fit on an Arduino Proto Shield. I did not work out every detail of connections, but I did run across the following potential parts and related information:
Camera (RS232 & TTL editions):
15nS SRAM 32Kx8:
In theory, the camera and SRAM would connect directly to the ARM CPU, on the shield, with a SPI connection to the Uno attached to the shield. The only real problem I ran into was when I started doing the math--32KB is not enough SRAM to handle a 320x240 image, and certainly not a 640x480 image. On top of that, I would prefer that there be enough RAM to support 2 full frames, so that the Uno can work on (process) 1 frame (from RAM) while the ARM stores the next frame (to RAM).
In order to get more RAM, we need a different physical package (not DIP), and if we go with surface-mount, then we might as well use a different package for the CPU, thus opening up a huge list of potential CPUs. The downside is that surface-mount is much more difficult to prototype, although it can be done (I know a guy who does it).
Anyway, my thought is that if you want this to be successful, and as flexible as you specify, then it will require a reasonably fast ARM CPU and a whole bunch of SDRAM connected to that CPU. It should be possible to store an entire frame in SDRAM, and let the Uno grab just the portion of the image that it wants (given X, Y coordinates). Plus, I like the flexibility of the image sizes and data formats supported by the camera. To make a long story short... I don't think an SD card is required, particularly if you want real-time image processing.
So, I would like to know if other (more) people would be interested enough in such a project that they might wish to help fund its development, with the assumption of receiving finished boards when it is all working. The biggest 2 expenses would be the cost of having some circuit board company actually populate the production boards, and the initial cost of a commercial-use (Standard) edition of Eagle CAD. I'm not sure the cheap 2-layer edition will suffice. Plus there is the expense of all the miscellaneous parts (CAM, SDRAM, CPU, connectors, etc.)
I believe I have enough knowledge (and/or people connections) about ARM, AVR, UART, SPI, C/C++, etc. to pull things together. I have a BS in Computer Science, and an MA in Digital Media (so I know something about video formats). With some playing, tweaking, and lots of prayer, it sounds doable in my mind.
But I don't have the cash needed to front the development. It's that simple.