Efficient recording and retrieval of complex digital Fresnel holograms based on the line partitioned autoassociative memory

In this paper, a novel method for the recording and retrieval of multiple digital Fresnel holograms, each corresponding a three dimensional (3D) object scene, is presented. As the hologram is complex (composing of a real and an imaginary parts), and its data size is generally larger than the optical image it represents, a classical AAM required to record the holographic data is enormous even for a medium size hologram. In view of this, we first convert the hologram into a binary format with error diffusion (a process hereafter refer to as 'binarization'). Each row of the hologram is recorded in a sub-autoassociative memory (SAAM), resulting in a network that is over 4 orders of magnitude smaller in size than the use of a single, classical AAM for recording the hologram directly. Our proposed AAM is referred to as the line partitioned autoassociative memory (LP-AAM). Experimental results demonstrate that our proposed LP-AAM is effective in retrieving a binary hologram when a corrupted or noise contaminated version of it is presented. Subsequently, the retrieved binary hologram can be taken to reconstruct the pictorial content with a quality that is comparable to that represented by the original hologram before binarization. To our knowledge, this is the first time an autoassociative memory is developed for the handling of holographic images.