Scientific instruments as payload of current space missions are often equipped with high-resolution sensors. Thereby, especially camera-based instruments produce a vast amount of data. To obtain the desired scientific information, this data usually is processed on ground. Due to the high distance of missions within the solar system, the data rate for downlink to the ground station is strictly limited. The volume of scientific relevant data is usually less compared to the obtained raw data. Therefore, processing already has to be carried out on-board the spacecraft. An example of such an instrument is the Polarimetric and Helioseismic Imager (PHI) on-board Solar Orbiter. For acquisition, storage and processing of images, the instrument is equipped with a Data Processing Module (DPM). It makes use of heterogeneous computing based on a dedicated LEON3 processor in combination with two reconfigurable Xilinx Virtex-4 Field-Programmable Gate Arrays (FPGAs). The thesis will provide an overview of the available space-grade processing components (processors and FPGAs) which fulfill the requirements of deepspace missions. It also presents existing processing platforms which are based upon a heterogeneous system combining processors and FPGAs. This also includes the DPM of the PHI instrument, whose architecture will be introduced in detail. As core contribution of this thesis, a framework will be presented which enables high-performance image processing on such hardware-based systems while retaining software-like flexibility. This framework mainly consists of a variety of modules for hardware acceleration which are integrated seamlessly into the data flow of the on-board software. Supplementary, it makes extensive use of the dynamic in-flight reconfigurability of the used Virtex-4 FPGAs. The flexibility of the presented framework is proven by means of multiple examples from within the image processing of the PHI instrument. The framework is analyzed with respect to processing performance as well as power consumption.