An image watermarking scheme with dual purpose is proposed by introducing the selfembedding technology, which can achieve both copyright protection and content authentication. At the embedding side, an original image is split into nonoverlapping blocks, and the m least significant bits of every pixel in each block are set to be zero. Then each block is conducted with singular value decomposition, and an original robust feature watermark is derived from judging the parity of the first digit of singular value′s norm from each block. Finally, a watermarked image is produced by selfembedding the original robust feature watermark into the m least significant bits of every pixel in each block. At the detection side, robust feature watermark′s extraction from an attacked watermarked image is similar to its production at the embedding end. Copyright identification is accomplished by calculating the normalized correlation between the extracted robust feature watermark and the original robust feature watermark. Content authentication is accomplished by tamper detection through judging the consistency between the extracted robust feature watermark and the m least significant bits of every pixel in each block of attacked image. Both theoretic analysis and experimental results illustrate that it has perfect invisibility. Experimental results also show that it not only has strong robustness towards attacks such as adding noise, cropping, JPEG compression, smoothing, resampling and geometric attacks like random row removal, right shifting and downward shifting, but also can accurately locate the tampered region and distinguish tamper type. Therefore, it can achieve dual purpose containing copyright protection and content authentication.