A spot centroiding algorithm immune from noise pixels is proposed to improve the existing centroiding algorithms that can not process images with discrete noise pixels directly, and it is achieved by a Field Programmable Gate Array(FPGA). This algorithm does not depend on noise pre-process or post-process to eliminate noise pixels. Firstly, the algorithm marks the background pixel, noise pixel and the spot pixel, respectively, then each pixel is marked after it is compared with the neighbouring one. At the same time, centroid parameters that belong to the same spot are accumulated, while the real noise pixel is bypassed. Compared with currently existing algorithms, this method takes full advantage of the parallel processing ability of the FPGA. It can extract spot centroid coordinates and eliminate the noise pixel simultaneously when the image pixel is output, the pre-process image is not needed and the storage space is saved. This algorithm is especially suitable for processing the spot image with discrete and high-brightness noises caused by long exposure time.