Analysis of DNN based image watermarking data generation for self-recovery

Abstract

The world is rapidly going digital and data security is the foremost concern. Digital images form a large chunk of this digital data. Therefore, techniques like digital image watermarking are used for assuring image security. In this paper application of DNN based compression algorithm is examined for implementation of color image tampering detection and self-recovery system. The DNN based image compression is used to compress the host images and multiple copies of compressed images are embedded in the host image for self-recovery. Due to the smaller size and better quality provided by DNN based compression algorithm, the watermarked images have high PSNR (peak signal to noise ratio). Similarly, the quality of recovered images is also high. The DNN based algorithm is analyzed for different images and tampering conditions. In the best-case scenario, the algorithm can recover the image from about 88% tampering.