Abstract: Fingerprint image filtering is the most time-consuming process of fingerprint image enhancement. In order to improve the real-time performance of fingerprint filtering, a Separable Gabor filter is proposed. The two-dimensional Gabor filter is split into a one-dimensional band-pass filter and a one-dimensional low-pass filter. Two filters respectively convolute the fingerprint image in mutually perpendicular directions. The algorithm is realized on FPGA using hardware description language. In order to save memory space in the implementation process, coordinate transformation and direction transformation were introduced. Experimental results show that for the same fingerprint image, the processing results of the Separable Gabor filter and the Gabor filter are similar, and the former is 1.9 to 4 times faster than the latter, the consumption of memory space of the Separable Gabor filter is only 16.4% of the Gabor filter Increases the filtering speed while greatly reducing hardware costs.