Fixed threshold algorithm can achieve better denoising effect due to the small spectral coefficients caused by noise being set to zero. But because of its threshold function is not continuous and a mutation in the vicinity of the threshold, so some big noise spectral coefficients′ reservations and small spectral coefficients result in the loss of information. In order to improve filtering algorithm threshold for windowed Fourier transform method, an adaptive threshold method is proposed. Through the verification of simulated speckle pattern interferometer phase map, the adaptive threshold method can effectively filter out the noise with the signal information well retained.