Depth consistency and vertical disparities in stereoscopic panoramas

Video in TIB AV-Portal: Depth consistency and vertical disparities in stereoscopic panoramas

Formal Metadata

Depth consistency and vertical disparities in stereoscopic panoramas
Title of Series
Part Number
Number of Parts
CC Attribution - NoDerivatives 2.0 UK: England & Wales:
You are free to use, copy, distribute and transmit the work or content in unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor.
Release Date

Content Metadata

Subject Area
CONTEXT: In recent years, the problem of acquiring omnidirectional stereoscopic imagery of dynamic scenes has gained commercial interest and, consequently, new techniques have been proposed to address this problem [1]. The goal of many of these novel panoramic methods is to provide practical solutions for acquiring real-time omnidirectional stereoscopic imagery suitable to stimulate binocular human stereopsis in any gazing direction [2][3]. In particular, methods based on the acquisition of partially overlapped stereoscopic snapshots of the scene are the most attractive for real-time omnistereoscopic capture [1]. However, there is a need to rigorously model these acquisition techniques in order to provide useful design constraints for the corresponding omnidirectional stereoscopic systems. OBJECTIVE: Our main goal in this work is to propose an omnidirectional camera model, which is sufficiently flexible to describe a variety of omnistereoscopic camera configurations. We have developed a projective camera model suitable to describe a range of omnistereoscopic camera configurations and usable to determine constraints relevant to the design of omnistereoscopic acquisition systems. In addition, we applied our camera model to estimate the system constraints for the rendering approach based on mosaicking partially overlapped stereoscopic snapshots of the scene. METHOD: First, we grouped the possible stereoscopic panoramic methods, suitable to produce horizontal stereo for human viewing in every azimuthal direction, into four camera configurations. Then, we propose an omnistereoscopic camera model based on projective geometry which is suitable for describing each of the four camera configurations. Finally, we applied this model to obtain expressions for the horizontal and vertical disparity errors encountered when creating a stereoscopic panorama by mosaicking partial stereoscopic snapshots of the scene. RESULTS: We simulated the parameters of interest using the proposed geometric model combined with a ray tracing approach for each camera model. From these simulations, we extracted conclusions that can be used in the design of omnistereoscopic cameras for the acquisition of dynamic scenes. One important parameter used to contrast different camera configurations is the minimum distance to the scene to provide a continuous perception of depth in any gazing direction after mosaicking partial stereoscopic views. The other important contribution is to characterize the vertical disparities that cause ghosting at the stitching boundaries between mosaics. In the simulation, we studied the effect of the field-of-view of the lenses, and the pixel size and dimension of the sensor in the design of the system. NOVELTY: The main contribution of this work is to provide a tractable method for analyzing multiple camera configurations intended for omnistereoscopic imaging. In addition, we estimated and compared the system constraints to attain a continuous depth perception in all azimuth directions. Also important for the rendering process, we characterized mathematically the vertical disparities that would affect the mosaicking process in each omnistereoscopic configuration. This work complements and extends our previous work in stereoscopic panoramas acquisition [1][2][3] by proposing a mathematical framework to contrast different omnistereoscopic image acquisition strategies.