Mechanism to capture and communicate image-processing expertise

The problems that algorithm designers encounter when existing programming environments and languages are used to solve real-world image-processing tasks are discussed. A method is proposed for abstracting, storing, and sharing image-processing knowledge that involves decomposing complex tasks into primitives (atoms) and then integrating them into new functional tools (molecules). The discussion covers operator classifications; the mechanism object, which makes it easier to generalize transformations (by conversion operators) representing complex associations between operators and transforming operator graphs; and building mechanisms.<
>