For the second generation of digital watermarking technology, this paper proposes a DWT image watermarking algorithm based on Harris feature points and singular value decomposition. First the Harris detector was utilized to extract steady feature points from the standard image. Then wavelet decomposition was conducted for the local feature regions to obtain low frequency sub-bands. These sub-bands were then divided into blocks. Singular value decomposition was conducted for each block. The watermark information was embedded by weighing the maximal singular value in each block. PSNR was not less than 45 dB and NC value closed to 1, indicating that the method was feasible. In conclusion, the algorithm has very good robustness to shear attacks. Meanwhile it can resist common signal processing attacks, such as noise, median filter, improved the brightness, reduced the brightness, and JPEG compression.