An auto-adaptive scheme of embedding watermark based on wavelet transformation domain is proposed in this paper. Firstly, the scheme applies wavelet transform to the original image, confirms the ideal embedded location in low-frequency coefficient of wavelet transformation domain by using FCM algorithm integrating with the apperceive characteristics of human visual system and related local characteristics of image. According to quantization rule of wavelet coefficient proposed in this paper, the information of scrambled watermarking is adaptively embedded in corresponding position. The method of finding the embedded location in the process of watermark extracting is same as the method of embedding watermark. Secondly, extract the information of embedded watermark according to the rule of coefficient quantization. Finally, after the process of anti-scrambling and enhancing dimension, the embedded watermark image is obtained. Experimental results show that the proposed scheme has performance of good invisibility and robustness.