Invisible QR Code Generator Using Convolutional Neural Network

Abstract

The authors aim to embed arbitrary information in arbitrary images and restore them using CNN. To achieve this goal, we propose a model consisting of two CNNs with different roles. In the proposed method, it is used as an information medium for embedding a QR code. The QR code error correction function is expected to restore the embedded information without error. Existing research has shown that embedding a QR code in a sharp color image does not restore the QR code correctly. This paper modified the CNN configuration to address this issue. The authors hope this technology can be used to integrate QR codes into human living space and hide information without upset. We learned how to embed a QR code image in a color image using the CNN model proposed this time. As a result, the authors were able to embed the QR code image without degrading the quality of the input color image. Current methods have drawbacks. Blur the image with the embedded QR code. Then, there is a problem that the embedded QR code cannot be restored. We will solve this problem in the future.