Light field image encryption based on steerable cosine number transform

Abstract

Light field (LF) images can be represented in many different ways but the most convenient, in terms of visualization and processing, seems to be as an array of 2D images, that is, a 4D structure that displays the spatial and angular planes of the image. Previously published works do not consider the entire 4D structure. They apply the encryption technique only in the spatial plane or in both planes separately, which maintains a certain correlation between neighboring pixels. The present paper proposes an encryption technique for LF images considering the encryption of both planes at once. The main motivation for this approach is to achieve a lower correlation coefficient, thus being more robust against statistical attacks. We introduce the steerable cosine number transform (SCNT), which is obtained by rotating the base vectors of the two-dimensional cosine number transform (2D-CNT), using a finite field rotation operator, and we propose an encryption method for light field images based on the SCNT, which uses the angles of the referred rotations as secret parameters. Simulation results show that the proposed scheme is robust against most cryptographic attacks and highly sensitive to changes in the secret key.