Image processing is an area with many computat ionally demanding algorithms. When implementing an algorithm the programmer has to make the choice of either using a high-level language, thereby gaining rapid development at the expense of run-time pe rformance. Or, using a lower level language, having higher run-time performance but also a higher implementation cost. In th is paper we present PyGPU, an embedded language that enables image processing al gorithms to be written in the high-level, object- oriented language Python. PyGPU functions are comp iled to execute on the graphics processing unit (GPU) present on modern graphics cards, a streaming processor capable of speeds more than a magnitude higher than those of current generation CPUs. We demonstrate a number of common image processing algorithms, showing how th ese can be implemented succinctly and clearly using high-level abstractions, while at the same time achieving performance close to theoretical peak figures.