We discuss the use of Field Programmable Gate Arrays (FPGAs) as hardware accelerators in genetic algorithm (GA) applications. The research is particularly focused on image processing optimization problems where fitness evaluation is computationally demanding and poorly suited to micro-processor systems. This research identifies key design principles for FPGA based GA and suggests a novel 2 stage reconfiguration technique. We demonstrate its effectiveness in obtaining significant speed-up; and illustrate the unique hardware GA design environment where representation is driven by a combination of hardware architecture and problem domain.