A Bayesian Approach for Segmentation in Stereo Image Sequences
- PDF / 1,052,942 Bytes
- 11 Pages / 612 x 792 pts (letter) Page_size
- 43 Downloads / 190 Views
A Bayesian Approach for Segmentation in Stereo Image Sequences George A. Triantafylllidis Information Processing Laboratory, Electrical and Computer Engineering Department, Aristotle University of Thessaloniki, Thessaloniki 54006, Greece Email: [email protected]
Dimitrios Tzovaras Informatics and Telematics Institute, 1st Km Thermi-Panorama Road, Thermi-Thessaloniki 57001, Greece Email: [email protected]
Michael G. Strintzis Information Processing Laboratory, Electrical and Computer Engineering Department, Aristotle University of Thessaloniki, Thessaloniki 54006, Greece Informatics and Telematics Institute, 1st Km Thermi-Panorama Road, Thermi-Thessaloniki 57001, Greece Email: [email protected] Received 31 August 2001 and in revised form 14 May 2002 Stereoscopic image sequence processing has been the focus of considerable attention in recent literature for videoconference applications. A novel Bayesian scheme is proposed in this paper, for the segmentation of a noisy stereoscopic image sequence. More specifically, occlusions and visible foreground and background regions are detected between the left and the right frame while the uncovered-background areas are identified between two successive frames of the sequence. Combined hypotheses are used for the formulation of the Bayes decision rule which employs a single intensity-difference measurement at each pixel. Experimental results illustrating the performance of the proposed technique are presented and evaluated in videoconference applications. Keywords and phrases: Bayesian decision test, segmentation, stereoscopic video, disparity, motion.
1.
INTRODUCTION
Stereo vision provides a direct way of inferring the depth information and therefore 3D perception by using two images (a stereo pair) destined for the left and right eye, respectively. Stereoscopic video sequences are represented by a number of consecutive stereo pairs each of which corresponds to a constant time instant. In a stereoscopic image sequence, each different view in a stereo pair is recorded with a difference in the observation angle, creating an enhanced 3D feeling to the observer and increased tele-presence, for example, in teleconferencing [1]. Such video can provide more vivid and accurate information about the scene structure than monoview video. For these reasons, stereoscopic video processing has been the focus of considerable attention in recent literature [2, 3, 4]. As stereo vision is used in more and more applications, efficiency in stereo sequence coding, transmission and storage becomes increasingly important. The major goal is to
exploit the correlation between the two frames (left and right) of a stereo sequence, along with the correlation which is present between the stereo pairs at consecutive time instances, in order to achieve reduction of the number of bits required to represent the stereo sequence [1, 4, 5]. Increased data compression will make the stereo imaging applications faster and more economical. Accurate detection of occlusion and visible foreground and background areas in s
Data Loading...
