Development of a Stereoscopic Image Processing Software
Nobuhiko Mori
Earth Remote Sensing Data Analysis Center (ERSDAC)
Forefront Tower, 3-12-1, Kachidoki, Chou-ku, Tokyo 104, Japan
Abstract
A digital stereoscopic image processing software has been developed on a personal computer (PC) system. The system was originally developed as a digital photogrametric system, and had a special three dimensional (3d) display module which could display stereoscopic drawings together with stereoscopic images. The stereoscopic drawings were used as a tool of man-machine interface. For instance, stereoscopic points on stereoscopic images were used to inform the positions of some stereo conjugate points to computer, and to correct the results of computer stereo matching. The functions of the stereoscopic image processing software which have been developed on this system are mostly similar to those of an ordinary two dimensional image processing software. Using the software, images can be processed, overlaid and changed into stereoscopic images. Further more, this stereoscopic graphics can be used as a tool of man-machine interface. This ability has been taken over from the digital photogrametric system. Thus software can be used widely where geographical information is necessary. As an application example, stereoscopic images which can be used in earth resource exploration field have been made using this software and presented here.
1. Introduction
Topographic information is very important in many application fields, such as earth resource exploration, route selection (road, pipeline, etc.), dam construction and wide area development. However, it is very hard to get topographic information on digital systems because usually they have only two dimensional display. some digital photogrametric system have 3D display, but no sufficient image processing software attached. In this work, a stereoscopic image processing software has been developed on a PC system with 3D display. This system was originally developed as a digital photogrametric system, but now the ability of a digital stereoscopic image processing system is developed.
This system has a special 3D display module, which has an ability to display stereoscopic drawings together with stereoscopic images. Figure 1 shows the function of the 3D display module. In figure 1, stereoscopic drawings are memorized in Video Random Access Memories (VRAM), and stereoscopic images are in Frame Memories. Left image and drawing, and right image and drawing are overlaid respectively and displayed on a display one by one alternately. Observer can get a stereoscopic view by using liquid crystal shutter glasses, which are synchronized with the display. By using this module, this system get a special ability as man-machine interface.
Figure 1 The function of the 3D display module
2. Man-machine interface in DEM extraction from digital stereoscopic images
Generally speaking, image recognition is very difficult for a computer because there are too many informations in images to be analysed. It is also true in the case of stereo matching, which is one technique to determine stereo conjugate points automatically with digital photogrametric systems. Now the errors of stereo matching are very serious problem for automatical DEM extraction, because it is impossible to get rid of them completely by present computer techniques. Many methods are already proposed, but they all need to be improved. It seems that the best way is to get human help at appropriate stages. In order to get human help, man-machine interface (3D display module) is necessary. In automatical stereo matching, man-machine interface can be used to reduce matching errors at two stages, before and after the stereo matching.
Before the stereo matching, some stereo conjugate points can be informed to a computer by using the man-machine interface. Photo 1 is one example. In this case, 25 stereo conjugate points were informed to the computer, which were determined manually by using 3D display module. Because this module can display stereoscopic images together with stereoscopic drawings of points, the hight of the points can be easily adjusted to the hight of earth surface of images by using a mouse device.
Photo 1 Manually determined stereo conjugate points
Photo 2 is another example. In this case, stereo conjugate lines were informed to the computer. These lines were drawn one by one along the earth surface of images with a mouse looking at the 3D image through liquid crystal shutter glasses. The sequence to draw these lines are as follows.
Photo 2 Manually drawn stereopair lines
-
First draw one line of stereo pair along a valley or a ridge in one (left or right) image.
- Next draw the second line just on the line mentioned above in the other image.
- Repeat these two kinds of drawings.
These lines are very useful for automatical stereo matching because the accuracy of stereo matching is usually bad at valleys and ridges when a small area correlation method is used.
After the stereo matching, the errors of automatically determined stereo conjugate points can be corrected by using the man-machine interface. Photo 3 is a simultaneous display of stereo-scopic images and stereo conjugate points determined automatically. When observed through liquid crystal shutter glasses, correct points are positioned on the surface of a model of a mountain and error points are positioned in the air or under the ground. These error points can be corrected by the mouse easily. In the photo, a pair of big cross marks indicate the points under correcting.
Photo 3 : Simultanious display of stereoscopic images and stereo conjugate points.
