Abstract: The infrared image has disadvantages, such as the blurred image and the strong noise. In order to get better effects of detection and recognition, image denoising is an important work. The basic principle of wavelet transformation was presented, and it was used to process the infrared image of defects, which was connected with the median filtering and the principle component analysis. The new method did not rely on accurate estimation of noise variance. The experimental results showed that the processing effects were better than traditional methods, and it could effectively reduce the Gaussian and impulse noise at the same time. Presented method could therfore be used for the further analysis and processing of the defects.