Lucky imaging techniques implemented by the FastCam group (see http://www.iac.es/proyecto/fastcam/) at the Instituto de Astrofisica de Canarias have demonstrated its ability to obtain spectacular diffraction limited images in telescopes ranging from 1 to 4.2 m in visible wavelengths (mainly in the I band), at the expense of using only a small percentage of the available images. This work presents the development of a real-time processor, FPGA-based, capable of performing all the required processing involved in the lucky imaging technique: Bias and flat-field correction, quality evaluation of images, quality threshold for image selection, image recentering and accumulation, and finally sending through Gigabit Ethernet both raw and processed images to a PC computer. Furthermore, a real time display is generated directly from FPGA showing both types of images, plus a histogram of the computed quality values and the threshold used. All processes can co-exist physically located in separated places inside the FPGA, using its natural parallel approach, and can easily handle the 512x512 pixels at 30 fps found at the sensor camera output (an Andor Ixon+ DU-897ECSO EMCCD). Flexibility and parallel processing features of the reconfigurable logic have been used to implement a novel imaging strategy for segmented-mirror telescopes, allowing separate evaluation of every segment and posterior accumulation to achieve the resolution limit of a single segment with the integration capability of the full primary mirror.