[4I3-GS-7] Language media processing:

[4I3-GS-7-04] Designing a Coded Aperture for Visual Information Hidden Using Evolutionary Multi-objective Optimization

〇Yasura Yokogawa¹, Kazuhiro Ohta¹, Tomoki Minamata¹, Hiroshi Kawasaki², Hajime Nagahara³, Satoshi Ono¹ (1. Kagoshima University, 2. Kyushu University, 3. Osaka University)

[[Online]]

Keywords:Computational photography, Coded Aperture, Information Hidden, genetic algorithm

Coded aperture is a technique that replaces a normal circular camera aperture with a complex two-dimensional pattern. The coded aperture technique requires to design of a specific aperture shape according to a target task, such as deblurring, depth estimation, and so on. This study proposes a method for designing a coded aperture for information hiding. The proposed method adopts a multi-objective genetic algorithm to simultaneously design an aperture shape suitable for information hiding and a perturbation pattern to hide secret information. The secret image is hidden by a cover image that looks like random dots, and the perturbation and inverse convolution are applied to minimize the visibility of the secret information at the focal distance. This makes it possible to decode the secret information only when the secreted image is captured using a specific aperture (key aperture) at a shooting distance outside the focal distance. The decoding accuracy is minimized for apertures other than the key aperture. Experimental results show that it is possible to design a coded aperture that makes it difficult to see the hidden information at the focal distance and that enables decoding of the hidden information at a specific distance outside the focal range.