Modular architecture for real-time astronomical image processing with FPGA

Abstract

This paper describes the implementation of a two-dimensional, real-time centroiding algorithm using reconfigurable Field Programmable Gate Arrays (FPGA's). A centroiding computation is used to calculate the highest image light density location to sub-pixel resolution for bright point source events. The method is of interest mainly in astronomical applications. In order to achieve this computation in real-time, a solution which realizes high parallelism, pipelinability, and modularity is essential. Recent improvements in FPGA architecture allow for a new approach to realizing fast, wide and complex arithmetic functions thereby targeting these devices for real time imaging applications. A prototype has, therefore, been developed exploiting the high density, high speed and flexibility of FPGA's. The resulting prototype is able to perform centroiding on a 5×5 pixel array on data coming from an electronic imaging detector at a throughput rate of 10Mevents/s.