To enhance the performance of image authentication, a new fragile watermarking scheme, which exploits the perturbation in reverse processing, is proposed. In verifying the integrity of image contents, the method performs the reverse processing of watermarking. Typically, it de-filters the distributed version or solves an embedding equation instead of really extracting the watermark. If any tampering happened, the output should be perturbed violently because such processing enlarges the observation error, which can be regarded as the consequence of illegal manipulation. The drastically perturbed values imply the existence of tampering, and their positions directly draw the shapes of the manipulated areas. Compared with the mostly used block-based watermarking, the method localizes the tampering almost pixel-wise. It also supports the adaptive embedding, which keeps the perceptual quality better, and avoids the vulnerabilities resulting from the block-based approaches.