Aiming at the problems of the traditional ghost imaging reconstruction algorithm with many sampling numbers and poor reconstructed image quality,a di fferential ghost imaging optimization method based on speckle image filtering is proposed.In the correlation stage of differential ghost imaging,this method f irst performs corresponding filtering processing on the random speckle pattern; then associates the original bucket detector value with the processed speckle pa ttern to obtain the reconstructed ghost image after noise reduction.Finally,th e detailed information of the reconstructed image is supplemented by the method of image fusion.Numerical simulation results show that compared with traditiona l differential ghost imaging,the peak signal-to-noise ratio of the reconstruc te d differential image in this paper is improved by 2-4dB under the same sampling numbers.Experimental results show that the algorithm can take into account the contradiction between sampling numbers and image quality.It can also achieve go od recovery effects for ghost imaging under low sampling numbers.