The authors propose an image authentication scheme, which is able to detect malicious tampering while tolerating some incidental distortions. By modeling the magnitude changes caused by incidental distortion and malicious tampering as Gaussian distributions with small and large variances, respectively, they propose to embed a watermark by using a mean-quantization technique in the wavelet domain. The proposed scheme is superior to the conventional quantization-based approaches in credibility of authentication. Statistical analysis is conducted to show that the probabilities of watermark errors caused by malicious tampering and incidental distortion will be, respectively, maximized and minimized when the new scheme is applied. Experimental results demonstrate that the credibility of the method is superior to that of the conventional quantization-based methods under malicious attack followed by an incidental modification, such as JPEG compression, sharpening or blurring.