Excessive noise was joined with the acquiring wood defect images process in wood non-destructive testing. The image quality was affected. Wavelet packet transform could decompose high frequency and low frequency of signal at the same time. There were characteristics of open and closed in Gray-scale Morphology. Denoising method of wood defect images based on wavelet packet combined with gray-scale morphological filtering was proposed. Compared with the traditional denoising method, the method received better Peak Signal to Noise Ratio (PSNR) and visual effect. Experimental results verified the feasibility and effectiveness of this method.