Summary: This article describes an approach for automatically generating efficient code for region-based binary image processing algorithms. Inspired from techniques used in digital system design, this approach can be applied to any binary image processing algorithm which evaluates a discrete function over small regions of the image (e.g., discrete morphological and topological operations). Given a description of the function in terms of either Boolean formulae or a compiled module, it automatically produces a program which implements the function. The generated C source code is portable and compact. It is also very efficient: at each stage of its execution, the procedure is guaranteed to examine only the pertinent input data, i.e., the values which affect the result. For each such value, it performs at most one test, one branching, and one binary register operation. We apply our method to several binary image processing tasks, such the 2D and 3D simplicity tests, and two different 3D thinning processes. In each case, we produce functions more efficient than the previously optimized implementations, reducing the execution time by a factor of up to 20.