A Color Image Authentication Scheme With Grayscale Invariance

Color image authentication is a method that allows the detection of tampered regions on images. Existing related works evaluated their performance based on visual quality and detection capability of the marked image, but the issue of grayscale invariance was ignored. However, many applications in image processing required the color image to first convert into a grayscale image before any further post-processing such as edge detection, color masking in Photoshop, and the display of e-ink. If the resulting image and the original image are different in their grayscale value, they might produce different outcomes after the post-processing. Therefore, how to maintain unchanged of the grayscale value has become an important issue. This paper presents a grayscale-invariance color image authentication technique. We proposed to embed the authentication code into two of three primary color channels and adjust the remaining one to remedy the distortion of grayscale value. The experimental results showed that, when embedding each four-bit authentication code into two color channels, the visual quality of the marked images has achieved an average 33.26 dB of peak signal-to-noise ratio (PSNR) while providing a satisfactory detectability.