Peter Geißler

Three-dimensional imaging algorithms

Publisher Summary Image acquisition always contracts the 3-D information of the scene to 2-D information of the image due to the projection on the 2-D image plane. Therefore, the reconstruction of the depth information from 2-D images is a fundamental problem in computer vision applications. Many different approaches to the problem are known, including stereo or its generalization to multiview imaging, shape-from-shading and photogrammetric stereo, and shape from motion, texture analysis and depth from focus. This chapter focuses on two important approaches for the recovery of depth information: stereo reconstruction and depth-from-focus. Although not immediately apparent, these two methods are based on the same principle. Stereo uses the difference between images taken from different viewpoints, the so-called parallactic difference. The lateral shift of image points between the two images is directly correlated to the distance of the object point. Depth-from-focus uses the inherent blur in the images to correlate it with depth. Because blur cannot be directly separated from the image, further images that differ only in the grade of optical blur are required. So both methods gain the 3-D information from the difference of images taken from the same scene, but with different camera parameters in a general sense. In this sense both techniques have to be classified as a triangulation technique.

3 – Imaging Optics

Publisher Summary This chapter discusses computer vision. Image processing always starts with image acquisition, mostly done by illuminating the scene with natural or artificial light in the visible range and capturing images with a photographic lens. The importance of proper image acquisition is ignored in many applications, at the expense of an increased effort in the processing of the images. Appropriate visualization can enhance image quality in such a manner that image processing requires fewer processing steps, becomes much faster, or is even for the first time possible. Image degradations caused by unsuitable imaging may seriously complicate image analysis or even be uncorrectable afterwards. Although most of todays camera lenses are of very good quality, they are always optimized for a particular purpose and may fail if used in other setups. In some applications an optics setup from one or two simple lenses may provide better image quality than stock lenses because the setup can be optimized exactly for that imaging problem. For these reasons, the chapter provides essential concepts of optical imaging, focusing on the geometric ray approximation, which are sufficient for most applications other than microscopic imaging. Special emphasis is placed on the description of nonparaxial optics.

Segmentierung von Partikelbildern in der Strömungsvisualisierung

Eine der Hauptschwierigkeiten in der Stromungsvisualisierung besteht in der Segmentierung von Teilchenbildern, wie sie fur die Particle-Tracking-Velocimetrie benotigt werden. In diesem Beitrag wird ein Verfahren vorgestellt, das Tracerteilchen anhand der Lokalen Orientierung identifiziert und anschliesend mittels eines Fits an eine Modellfunktion deren Position und Bewegungsrichtung genau bestimmt. Damit ist es moglich, selbst Teilchen mit hoher Geschwindigkeiten, die in den Bildern als Streifen visualisiert werden, zuverlassig zu segmentieren.

In Situ Determination of Cell Concentration in Bioreactors with a new Depth from Focus Technique

A new depth-from-focus technique is introduced that requires only a single image to determine the distance of simple shaped objects from the focal plane. The technique has been applied to evaluate the concentration of cells in a bioreactor during a fermentation process. Since the low-intensity fluorescent light gathered by a light-amplifying camera results in images of low signal-to-noise ratio, an adaptive smoothing filter is used. A sharpness criterion derived from bandpass decomposition of the image in a Laplacian pyramid is used to define a virtual measuring volume. In this volume, process parameters such as cell concentration, cell size and intensity of cell fluorescence are evaluated. The technique is also suitable for other types of simple objects.