Abstract: With the development of watermarking technology, most of the watermarking methods can effectively resist the general attacks, but there is still great difficulty for geometric attacks and combined geometric attacks while maintaining a high embedding capacity. Therefore this paper proposes a method which effectively resists geometric attacks and combined attacks. Firstly, the watermark is embedded into the R matrix the QR domain of image with strong robustness. Secondly, the feature parameters of the attacked image are extracted in the wavelet domain and attack parameters of image are preliminarily estimated using LSSVM, then attack parameters are estimated accurately by template watermarking. Lastly, the image is corrected geometrically and watermark is extracted. Meanwhile, in order to reduce the bit error rate of watermark, we also introduce BCH encoding technology. The theoretical analyses and experiments show that the proposed method has good robustness to the general geometrical attacks and combined attacks, and the embedding capacity of watermark is high.