As common as imaging operations are, the literature contains little about how to build systems for image computation. This paper presents a system which addresses the major issues of image computing. The system includes an algorithm for performing imaging operations which guarantees that we only compute those regions of the image that will affect the result. The paper also discusses several other issues critical when creating a flexible image computing environment and describes solutions for these problems in the context of our model. These issues include how one handles images of any resolution and how one works in arbitrary coordinate systems. It also includes a discussion of the standard memory models, a presentation of a new model, and a discussion of each one's advantages and disadvantages.