[22a-P04-1~11] 3.3 Information photonics and image engineering

[22a-P04-4] Structural similarity of reconstructed images due to different orthogonality of mask patterns in single-pixel imaging

〇Hiroki Takatsuka¹, Naoya Mukojima¹, Masaki Yasugi¹, Shiro Suyama¹, Hirotsugu Yamamoto¹ (1.Utsunomiya University)

Keywords:single-pixel-imaging, deep-learning, Hadamard

Single-pixel imaging can be used for privacy-aware imaging by capturing shadowgraphs. Therefore, we are working on gesture recognition with privacy in mind. However, a large number of measurements are required to obtain a clear image, and an efficient reduction of the number of masks has been proposed. In this study, we used a random dot pattern and an adamantine pattern as masks. In this study, we used random dot patterns and adamantine patterns as masks, because adamantine patterns have orthogonality and are expected to be efficient in reducing the number of masks.