I. P. Nikolaev

Hough Transform: Underestimated Tool In The Computer Vision Field

We discuss Hough Transform, some of its key properties, a scheme of fast and complete calculation of the Hough Transform (similar to the Fast Fourier Transform), and an efficient implementation of this scheme on SIMD processors. We also demonstrate an application of the Fast Hough Transform in computer vision by the example of an automatic page orientation detection unit incorporated in an intelligent character recognition system. Both 2D (scanner) and 3D (camera) cases of page acquisition are considered.

LCLV-based system for high resolution wavefront correction: phase knife as a feedback intensity producer

Abstract A nonlinear optical system with distributed feedback, based on a liquid-crystal light valve (LCLV), is considered. The use of a Fourier-filtering technique allows a feedback configuration which suppresses high-order aberrations. Predictions of the linear theoretical analysis are confirmed by computer simulations and experiments. Suppression efficiency characteristics, such as wavefront rms errors and Strehl ratios, are measured for distorted and corrected wavefronts. An outlook on the system performance improvement is discussed.

Advanced phase knife technique

Abstract Small-scale phase inhomogeneities visualization methods based on phase knife filtering are considered. A new modification of such a technique with controllable lower threshold frequency is suggested: when transversally displaced from the center of the Fourier plane, the phase knife demonstrated some new features. Phase-to-intensity transfer function of the filtering system for arbitrary depth of harmonical phase modulation is obtained analytically, numerically and experimentally. Possibility of optical implementation of “exclusive or” (XOR) logical operation by means of 2D phase knife is shown.

An optical feedback nonlinear system with a Takens-Bogdanov point: experimental investigation

Abstract A two-component nonlinear optical system based on a liquid crystal light valve (LCLV) with external feedback loops is experimentally investigated. A degenerate codimension-2 bifurcation point in the parameter space is experimentally located, and the nonlinear behavior of steady and drifting roll patterns in the vicinity of this point is investigated. The experimental results are found to correlate well with the nonlinear theory developed recently. The results reported represent, to our knowledge, the first experimental observation of such nonlinear effects in optics.

Spatiotemporal period doubling in a nonlinear interferometer with distributed optical feedback.

A single-pass nonlinear interferometer with feedback field rotation is considered. Increasing feedback gain results in excitation of spatial subharmonics that distort optical reverberators, which are basic output patterns of the system. Features of the period-doubling process depend on the number of the reverberators petals. A new spatiotemporal effect is observed: A basic structure with an odd number of petals is perturbed not by a static subharmonic but by a rotating wave.

The impact of speckle on the measurement of eye aberrations

We report the results of modeling the speckle structure of an optical field, which arises due to scattering of laser radiation at the retina of the human eye. The presence of such a speckle structure in the scattered radiation leaving the eye affects the accuracy of wavefront measurements aimed at determining the overall aberrations of the eye. We compared two methods for suppressing such speckle structures based on spatial and spectral averaging of a variety of speckle field realizations. The results of both theoretical calculations and numerical simulations appeared to be in a good agreement with experimental data obtained with the eye model and in vivo.

Evidence Maximization Technique for Training of Elastic Nets

This paper presents a technique of evidence maximization for automatic tuning of regularization parameters of elastic nets, which allows tuning many parameters simultaneously. This technique was applied to handwritten digit recognition. Experiments showed its ability to train either models with high accuracy of recognition or highly sparse models with reasonable accuracy.

Phase distortion suppression in a nonlinear optical system with integral feedback

We consider a mathematical model of an optical system where the phase modulation introduced by a nonlinear optical element involves a phase shift in an additional time-integral feedback loop, which is determined by the gain coefficient . In the case of , this model describes the dynamics of a nonlinear interferometer, where the effect of phase distortion compensation has been numerically and experimentally investigated before. In this paper, the results of 2D numerical simulations are presented for the case . It is shown that different types of small-amplitude stationary phase inhomogeneities can be fully compensated with time. The effect of the partial suppression of dynamic distortions given in the form of traveling waves and Kolmogorov-spectrum atmospheric irregularities is demonstrated; the quality of the distortion compensation is analyzed, depending on the spatiotemporal characteristics of the waves. The obtained results confirm the potential of using the optical feedback configuration with an additional integral circuit for suppression of both static and dynamic phase distortions.

Image differentiation with the aid of a phase knife

A scheme for optical differentiation of images based on spatial filtering is proposed. A phase knife—a glass plate consisting of two halves, whose thicknesses are chosen so that the optical path difference between the light waves passed through them makes up half the wavelength—serves as a filter. The linearity of the transformation and the sensitivity of the system to noise are investigated theoretically. The scheme was experimentally implemented on the basis of a liquid-crystal light valve. The experimental results confirm that such a setup can be efficiently used for processing images with a narrow-band spatial spectrum and for visualization of phase objects.

Depth-sensitive adaptive deconvolution of retinal images

The capability of resolving fine details on retinal images plays a key role in early diagnostic of vision loss. Biochemical and morphological features, which may be present in the early stages of many retinal diseases, cannot be detected today with current funduscopic instruments because of the losses in spatial resolution introduced by the ocular medium and cornea. One of the ways of the solution of such a problem is to use the adaptive optical systems first for measuring phase distortions and then for its suppression. In our work we suggest the innovative approach that includes two stages of adaptive correction. On the first stage a Shack-Hartman wavefront sensor and modal flexible mirror is used for low-order aberration correction. On the second stage a computer post-processing, or deconvolution, of the residual aberrations is done using the information on the aberrations measured by the sensor. In our report we present the specific design of the Shack-Hartman wavefront sensor suitable for measurements of human eye aberrations. The characteristics of a modal bimorph corrector are discussed. The features of deconvolution technique are outlined.