In order to fast and accurately extract the star centroid from high-resolution satellite imageries, this paper presents an algorithm-hardware design of real-time star centroid for deep space exploration missions. First, a block threshold segmentation method is proposed to eliminate the stray light effects in different regions. Then, an optimized one-scan centroid extraction algorithm with label combination is proposed for detecting complex star patterns and extracting its centroid. Moreover, we present an FPGA-based hardware architecture for the proposed algorithm. Experimental results show that the proposed algorithm-hardware design enables accurate and real-time star centroid extraction.