We present an efficient implementation of a state of the art algorithm PixelMatch for matching all pixels within consecutive video frames. The method is of practical interest for tracking movements in video; it is also related to blockmatching in standard video compression methods. From the source specification of PixelMatch, a limited number of high-level code transformations are first performed, and analyzed, to produce an intermediate executable software code. From the intermediate software code, an efficient reconfigurable hardware circuit is synthesized, in a fully automatic manner, to process Standard Definition video streams in real-time on a current mid-size FPGA. The software implementation compiled from the same code runs orders of magnitude faster than the original specification. Despite this, real-time software processing of video streams by PixelMatch is still only within reach of the highest-end workstations.