In semiconductor laser self-mixing interferometry (SMI) for micro-displacement measurement, the precise extraction of phase information is essential for high-accuracy displacement reconstruction. However, measurement noise induces phase errors in the SMI signal, leading to suboptimal displacement reconstruction accuracy. To tackle the challenge of signal denoising, wavelet thresholding denoising algorithms can effectively filter out most of the noise. However, they suffer from local oscillation issues when applied to SMI signal denoising. This results in the appearance of new interference peaks in the denoised self-mixing interference signal, thereby causing erroneous displacement reconstruction. This paper proposes an SMI signal processing algorithm that synergistically combines wavelet thresholding and Savitzky-Golay (S-G) filtering. By incorporating the S-G filtering algorithm, the algorithm smooths out noise at phase jump points on a global scale, thus mitigating the local oscillation issues inherent in wavelet-only denoising. Experimental results of displacement reconstruction indicate that the proposed method successfully eliminates high-frequency noise at both amplitude and phase jump points. Consequently, the reconstructed displacement curve retains the original waveform characteristics of the vibrating object.