Since the sole focus can't capture the high-quality image of multi-association plastic gear tooth profile's flaw detail on different end surface, these flaws couldn't be detected simultaneously. When it comes to the two images focused on different gear ends, the highly detailed image is required due to the large quantity and complex profile of gear teeth. In view of the subject above, some fusion rules of multi-focus image, which adopt big high-frequency coefficient and mean value of low-frequency coefficient, are proposed in this paper based on wavelet transformation. Then suitable wavelet coefficient and decomposition layers are adopted to fuse the multi-focus image. Experimental results show that the fusion image has a high quality as well as low calculation and the tooth profile is greatly different from the background. So the proposed method shows great foreground in real time detection.