The stochastic noise can be expanded on the basis of Fourier progression. It can be divided into two components, one is called ac noise which obeys the normal distribution statistically, on the other hand, the other is called dc noise which can decrease the detection SNR. Generally, a matched filter is used to reject noise, but it is only suitable for the rejection of white noise. In this paper, a wavelet filter is introduced to reject the dc noise for the signal detection of the pulse laser radar. The bubble function is selected as the mother function. The simulation results indicated that the SNR can be improved by wavelet filter when the dc noise is existing. It is indicated that the SNR is higher with matched filter than that with wavelet filter in the absence of the dc noise, while the SNR is greatly higher with wavelet filter than that with matched filter when dc noise is in existence.