Current stereoscopic camera arrangements for capturing 3D images, primarily parallel and converging radial configurations, are based on the disparity induced by the distance between the left and right cameras. As the camera-focus distance increases, the disparity diminishes. When the size of the magnified image of a pixel in the image sensor of each camera equals the camera distance, no disparity is observed (left and right are the same). These conventional arrangements are disadvantageous for small imaging devices such as mobile phones and handheld computers in taking pictures from a distance because the size of the devices limits how far apart the cameras can be. Moreover, image depth is also reduced because it depends on disparity. Accordingly, stereo cameras mounted on small imaging devices are only useful for objects a short distance away, although high-resolution image sensors with very small pixels can alleviate these problems somewhat.1
To solve these issues, we have invented a new stereo-camera arrangement. The optical axes of the two cameras are neither in parallel nor converge to a point but rather diverge to an angle between the devices. The back extension of the axes crosses at a common point. The disparity between left and right is determined by the distance and angle between them. In other words, disparity persists even when the distance goes to zero. The combination of minimal camera distance and constant-angle disparity makes this arrangement a practical solution for small display devices. In addition, it provides a pair of stereo images with better depth even for objects that are far away. The camera setup is similar to that of two feed horns in the focal-plane array of the parabolic antenna in millimeter-wave imaging systems.2,3 Each feed horn in the array has its own viewing direction and is scanned simultaneously with others in the array to acquire an image frame with reasonable resolution. Because the scanning axes of the feed horns have a common pivotal point, the image-taking geometry is similar to that of our diverging-camera array. In fact, the stereoscopic-image characteristics of the arrangement do not appear to be very different from that of the converging system (i.e., radial type), although this is still being investigated owing to the number of different ways in which the left and right images can be combined.
Figure 1 shows the optical geometry of our new setup. The diverging angle can be adjusted for better depth sensing for a given object distance. Figure 2 shows a pair of stereo images taken with the camera distance at zero and the diverging angle at approximately 1.5°. The depth of field is impressive.
Figure 1. Optical geometry of diverging stereo-camera arrangement.
Figure 2. A stereo image pair obtained by the diverging stereo camera.
There are two ways of combining the image pair for display. The first is to match their image centers. The second centers the images geometrically at O (normal to the optical axis of the stereo camera: see Figure 1). Depth perception differs depending on the method, the center approach being better suited to display panels and the optical-axis technique for projection.
In summary, our converging radial stereo-imaging configuration is well suited to mobile devices both because it gets around the problem for small space posed by conventional stereo cameras and because the diverging angle guarantees disparity between the pair without regard to imaging distance. We are currently investigating further applications of the diverging stereoscopic camera for microscopes and telescopes.
School of Computing and Communication
Kyungsan, South Korea
Jung-Young Son is a chaired professor in the School of Computer and Communication Engineering and a SPIE Fellow.
Department of Advanced Technology Fusion
Seoul, South Korea
Kwang-Hoon Lee received his MS from Soonchunhyang University (South Korea) in 2002. He is now studying for his PhD. His research interests include general optics, stereoscopic and multiview 3D displays, and human factors in depth recognition.
4. J.-Y. Son, Y. Gruts, J.-H. Chun, Y.-J. Choi, J.-E. Bahn, V. I. Bobrinev, Distortion analysis in stereoscopic images, Opt. Eng. 41, no. 3, pp. 680-685, 2002. doi:10.1117/1.1431249