Implementation and use of an image processing algebra for programming massively parallel machines

The implementation and use of a machine-independent image-processing language, the AFATL (Air Force armament laboratory) image algebra, on a massively parallel machine is presented. The authors introduce the problem of specifying image-processing algorithms in a machine-dependent way, introduce the image algebra, provide an overview of how image algebra constructs are implemented in Connection Machine *lisp and provide examples of the use of image algebra for a variety of image-processing operations. The generality, level of portability, and efficiency of the existing implementation are discussed. Results show that the interpreted execution speed of image algebra code compares favorably with hand-coded versions of similar algorithms and that the implementation can be easily modified to add further functionality