Three-dimensional image authentication with double random phase encryption in one capture.

Area of Science:

• Optics and Photonics
• Computer Vision
• Information Security

Background:

• Traditional 3D authentication methods often require multiple viewpoints or complex data handling.
• Integral imaging offers a promising avenue for capturing 3D information but faces challenges in processing and authentication.
• The integration of deep learning presents opportunities to enhance the efficiency and simplicity of 3D authentication systems.

Purpose of the Study:

• To propose and validate a novel three-dimensional (3D) authentication method.
• To develop a simplified 3D authentication process using advanced imaging and deep learning techniques.
• To demonstrate the feasibility of authenticating 3D data from a single captured image.

Main Methods:

• Development of a 3D authentication system based on double random phase integral imaging.
• Application of two neural networks for estimating depth information and synthesizing missing viewpoints.
• Integration of geometric refocusing techniques to streamline the capture, transmission, and storage of 3D data.
• Utilizing a nonlinear correlation method for the final authentication verification.

Main Results:

• Successful implementation of a 3D authentication method using a single viewpoint image.
• Demonstration of deep learning's effectiveness in estimating depth and inpainting synthesized images.
• Validation of the simplified authentication process through experimental results.
• Confirmation of the method's efficacy via nonlinear correlation analysis.

Conclusions:

• The proposed double random phase integral imaging method offers a novel and efficient approach to 3D authentication.
• Deep learning and geometric refocusing significantly simplify the 3D authentication pipeline.
• The method is experimentally verified, proving its potential for secure 3D data handling.