： Improvement of accuracy by machine learning for digital image watermarking based on Green-Noise spectrum spread method

Abstract

In the digital watermarking method based on Green-Noise spectrum spread, the embedding amount increases when the block size is reduced, but the false positives increase. Therefore, we tried to improve the accuracy of the classifier by using machine learning. As a result, a high accuracy with a judgment correct answer rate of 99% or more was obtained for an image in which 256 characters (2048 bits) were embedded in a 512x512 image. Also, in the resistance evaluation, it was confirmed that the resistance was stronger than that of the conventional method.