A novel stationary multiwavelet transform (SMWT) method was proposed. The method possesses the good quality of both multiwavelet transform and stationary scalar wavelet on signal denoising. The mallat decompositing and reconstructing of 2D image based on SMWT was inferred. By thresholding the SMWT coefficients of noisy infrared image, the original image can be reconstructed. The simulation results show that the image denoising effect of this method has obvious superiority compared with scalar wavelet and general multiwavelet transform method. At the same time, the reconstructing image can preserve the characterisics of original image as many as possible.