Abstract: MORRPH is a general-purpose reconfigurable processing unit, primarily intended for real-time 2-dimensional image processing. Its flexible architecture allows it to be used for other applications including 1-dimensional signal processing, 2-dimensional cellular automata problems, and 3-dimensional image processing. The modular, open-ended (incompletely specified) architecture consists of a rectangular mesh of processing elements. Each processing element contains one socket for a field programmable gate array (FPGA) chip and one socket for support processing chips. The FPGA chips provide an array of logic resources, consisting of combinational logic functions, flip-flops, and interconnections. Empty support sockets allow several discrete IC chips to be tightly coupled to each FPGA, the type of IC is determined by the individual requirements of the task to be performed by the MORRPH board. This allows the support chips to implement functions that are not efficiently realized by the FPGAs. These memory, arithmetic, or processing support chips are specified when a particular application is compiled for the MORRPH board. If a single MORRPH board does not provide enough computational power for a particular application, several MORRPH boards can be interconnected to solve the task. The MORRPH architecture is implemented as an adapter The modular nature of the MORRPH architecture at both the chip and board level allows a very efficient implementation of a required machine vision task.