The laser displacement sensor which uses position sensitive detector (PSD) as position detector can realize the displacement measurement in small space, but the PSD is vulnerable to external environment. Therefore, there are lots of noises in the output signal which severely affect the accuracy of PSD position detection. For the purpose of effectively extracting the useful PSD position signal, this paper presents a wavelet threshold denoising technology to improve the accuracy of position detection of laser displacement sensor based on PSD. Based on the discrete wavelet transform, the original signal of the PSD can be decomposed into multiple scales and then be reconstructed with quantified wavelet coefficients which contain the noise by soft thresholding function. The result of wavelet denoising is evaluated by the experiments such as using different kinds of base wavelet, changing the decomposition scale and considering different amounts of signal samples. The experimental results indicate that, by selecting the appropriate combination algorithm of wavelet denoising, the denoised signal can maintain the mean value of the original signal and the standard deviation can be minimized to 2 μm or below, which significantly improves the position detection accuracy of the sensor and shorten the measurement period.