To avoid the limitation of wavelet thresholding and the calculation complexity of non-local means filtering when an image is denoised, a more effective wavelet image denoising method based on Non Local Means (NLM) is proposed. Firstly, multi-level wavelet decomposition is carried out for an image containing noises. Then, a new BayesShrink estimation threshold is used to implement thresholding processing of the sub-band coefficients so as to remove the high frequency noise. Finally, to further remove the noise, NLM processing is implemented in part low-level sub-bands. The experimental result shows that compared with the common wavelet threshold denoising and NLM filtering methods, this method can remove the noises in an infrared image more effectively and can obtain a higher Signal-to-Noise Ratio (SNR) and a lower Mean Square Error (MSE). Moreover, the method is relatively simple in calculation and can achieve excellent visual effect.