Structured light has become a widespread technique for the development of camera-based 3D sensors. The structured illumination provides texture to homogeneous objects and thus allows for the reliable determination of the disparity of each object point in a stereo-camera setting. Even a monocular 3D sensor is possible if the light projector has a fixed relative position to the camera and if the structured light is coded, i.e. the position within the whole light pattern can be reconstructed uniquely from a small local window of the pattern, the uniqueness window. Coded patterns with such a uniqueness property are called Perfect SubMaps (PSM). In our paper we focus on the design and evaluation of the subset of symmetric isolated binary toroidal PSMs (SIBTPSM) for structured light patterns, because of their beneficial properties with respect to the signal-to-noise ratio and the use with laser light sources and DOEs. We define several figures of merit that are relevant for the practical use of PSMs in a 3D sensor: the PSM size, the size of the uniqueness window, the Hamming distance, the density, and the homogeneity. We have created SIBTPSMs using our own dedicated algorithms and have designed and fabricated DOEs that produce these patterns with large fan angles of 61° × 47° when used with near-infrared diode lasers (λ = 830nm). We analyze the influence of these characteristics on the 3D measurement process by theory, simulations, and experiments. The patterns of publicly available DOEs based on SIBTPSMs are used for comparison and reference. Our results show that the PSM width, the uniqueness window size, the minimum and average Hamming distances, and the uniformity have strong impact on either speed or quality of the 3D reconstruction, whereas the point density and the PSM height are of minor importance.
To segment complex and versatile image data from different modalities it is almost impossible to achieve satisfying results without the consideration of contextual information. In this approach, image segmentation is regarded as a high- dimensional optimization task, that can be solved by stochastical methods like evolutionary algorithms (EA). Initially, the iterative algorithm is provided with a set of good-quality sample segmentations. An efficient EA-based learning strategy generates a segmentation for a given target image from the provided samples. This two-level process consists of a global image-based optimization whose convergence is enhanced by locally operating pixel-based Boltzmann processes which restrict the search space to reasonable subsets. The stochastic reconstruction extracts the relevant information from the samples in order to adapt it onto the current segmentation problem, which results in a consistent labeling for the target image. The algorithm works unsupervised, because the range of possible labels and their contextual interpretation is provided implicitly by the sample segmentations. To prove the usefulness of the method experimental results based on both, reproducible phantom images and physiological NMR scans are presented. Moreover, an analysis of the basic segmentation and convergence properties is provided.