Semi-Fragile Watermarking for Tamper Detection of Color and Gray Images Based On Lifting Wavelet Transform and Nearest Neighbor

Today, a large number of multimedia data are easily accessible via the Internet and can be changed or modified simply by using image processing tools. Digital watermarking is a developing technology to ensure and facilitate data validation and security and also protect the copyright of digital media. One of the applications of digital watermarking is authenticating and detecting tampered region. Digital watermarking methods have been proposed for detection, are able to authenticate and check the integrity of the received image by utilizing the information that is already embedded in the media. In this paper, a new method is proposed for the detection of tampered region in the color and grayscale images. In the proposed method, at first, a lifting wavelet transform is applied to the image. Afterward, the horizontal edges are divided into the non-overlapping small blocks and then, the binary watermark is embedded in blocks by quantitating the two maximum factors. To extract the watermark, a set of statistical parameters are calculated for each block and considered as training data for the learning algorithm. For extracting the maximum correlation watermark, the nearest neighbor algorithm is used. Finally, the obtained watermark will be compared with the original watermark and the tampered region will be detected. The experimental results show that the proposed method is capable of detecting tampering area with high accuracy under various attacks such as histogram equalization, Gaussian filter, salt and pepper noise, sharpening, compression, and even combination attacks.