To exceed the capacity limitation of the surface-recording method of current optical data storage, the third dimension is introduced with photorefractive materials. Photorefractive materials are suitable for three-dimensional data storage in conjunction with nonlinear optical systems such as the two-photon absorption process of the material for recording and the confocal laser-scanning system for reading. I will describe the systems and the materials for three-dimensional digital memory with the experimental results for read-only memory with photopolymer, erasable memory with a lithium niobate crystal and rewritable memory with photochromic organic materials. The comparison between photorefractive digital three-dimensional memory with conventional holographic three-dimensional memory and near-field memory is also discussed in terms of dynamic range, noise, recording density, and accessibility.

In an earlier work, we have developed a new principle of optical microscopic tomography, by which an inside optical density distribution of a 3-D biological sample can be visualized without physically slicing. However, since the system is angularly band-limited, the spatial resolution (especially, the longitudinal one) is poor. In this paper, we describe a numerical method to improve the resolution of the reconstructed 3-D image by constraining the reconstructed density to be nonnegative. The method requires much smaller computation complexity than the conventional quadratic programming when the number of variates (pixels or voxels) are large (in the case of 3-D imaging, for example, 100×100×100). Experimental results by the proposed method with real data of a biological sample are shown to demonstrate the effectiveness of this method.

We developed a new method of Prony's spectral estimation. This method incorporates with eigenvalue -eigenvector decomposition and overdetermined system approach to estimate spectral parameters of exponentials from noisy data without a priori information. In this paper, we present the theorical derivation of the proposed method and also show noise influence by computer simulations. The mean errors of the frequency and damping rate were less than 17% and 5%, respectively, when SNR was 2. As one of applications, this method will be useful for quantitative analysis of experimental data.