An Anti-cropping Watermarking Method for Facial Images Using Prediction and Weber Ratio Techniques

Abstract

When a digital facial image is published in any social network, often the facial part of the image is cropped and abused without the owner permission. It is extremely difficult to verify the owner, especially in cases where the embedded watermark does not cover the cropped face part. This paper proposes a digital image watermarking method that is robust against cropping attack in the spatial domain for protecting the copyright of facial images. The face region in an image is evaluated by using phase correlation technique for embedding the watermark. A binary watermark is defined to be embedded in the facial region. Pixels for embedding the watermark are predicted and altered based on the Weber ratio to control the strength of the watermark in order to make it imperceptible to the human eye. The watermark is extracted from the watermarked image that does not require the original image by subtracting the watermarked pixel from its predicted pixel. After this, the pieces of the extracted watermarks are verified to measure their similarity with the original watermark. The experimental results show that our method yielded 82 % similarity with the original watermark image, when it was attacked by the cropping attacks for any shapes, sizes, locations and the JPEG compression.