Image invisibility assessment for visually encrypted images
Akira Uchida, Masaaki Fujiyoshi, Sayaka Shiota, Hitoshi Kiya · 2014
A variety of perceptual encryption schemes have been studied to generate visually encrypted images for protecting visual secrecy and privacy while supporting ongoing innovation and growth in the applications of digital imagery. However, typical image quality metrics such as PSNR and SSIM are not suitable to evaluate the invisibility of visually encrypted images. To overcome this issue, this paper proposes a novel full-reference invisibility index, referred to as the Triple feature similarity index (TFSI), for visually encrypted images. The TFSI is computed through four steps: 1) edge extraction as preprocessing, 2) extracting the Triple feature based on the Trace transform, 3) normalization of resolution, and 4) TFSI is computed between the reference and visually encrypted images. Experimental results show that the TFSI measures the invisibility of visually encrypted images.