This study presents an authentication method with self-recovery ability. The hidden watermark in the proposed method not only detects and locates the tampered portion of an image, but also self-recovers the tampered portion. To increase the ability to recover from tampering, the recovery data are shared by using an (r, n) threshold scheme and then scattered all over the image. Experimental results indicate that the visual quality of the watermarked image is acceptable, and that the positions of the tampered portions can be identified. Most importantly, the self-recovery result is very competitive.