Options
1990
Conference Paper
Titel
Stereo matching based on a combination of simple features used for matching in temporal image sequences
Abstract
In image sequence understanding, crucial tasks are the extraction of information about the 3D-content and motions from the depicted scene. Further, the recognition of objects in the scene is necessary. For the first two tasks, an intermediate level of description are displacement vector fields (DVF) which describe in the stereo case the disparity between the objects depicted by the two cameras and in time sequences the displacement due to the motion of the objects in the scene and/or the camera itself. A large amount of work has been done for the estimation of DVFs. The methods can be divided into two categories, one based on gradients of the gray values of the image and the other on the extraction of appropriate features from the image. The feature method is preferred here, because it comprises the advantage that those features can be used subsequently for the description of objects in the images. This paper gives a common framework for the processing of temporal image sequences and s tereo pairs using the same type of features. Compared with most of the work done in stereo matching; in this paper no a priori knowlge about the scene and no camera calibration is used. The stereo image pairs used has been taken from a stereo image sequence taken under real condition for a car driving in a road scene.