Combined use of BP neural network and computational integral imaging reconstruction for optical multiple-image security

Abstract

Integral imaging can provide a feasible and efficient technique for multiple-image encoding system. The computational integral imaging reconstruction (CIIR) technique reconstructs a set of plane images along the output plane, whereas the resolution of the reconstructed images will degrade due to the partial occlusion of other reconstructed images. Meanwhile, CIIR is a pixel-overlapping reconstruction method, in which the superimposition causes the undesirable interference. To overcome these problems, we first utilize the block matching algorithm to eliminate the occlusion-disturbance and introduce the back-propagation neural network algorithm to compensate for the low-resolution image. In the encryption, a computational integral imaging pickup technique is employed to record the multiple-image simultaneously to form an elemental image array (EIA). The EIA is then encrypted by combining the use of maximum length cellular automata (CA) and the double random phase encoding algorithm. Some numerical simulations have been made to demonstrate the performance of this encryption algorithm.