Dirk Gandolph

Confidence evaluation for robust, fast-converging disparity map refinement

To improve upon the initial disparity estimates stemming from a local correspondence method, a subsequent refinement step is commonly employed. The performance of the stereo matcher used for the initial estimation inherently depends on the underlying image content. For some regions of an image, it may be difficult or even physically impossible to establish accurate point correspondences. This results in disparity estimates of varying accuracy and reliability. In this paper, a confidence map is proposed which combines the consistency and quality of a match. It explicitly models the reliability of each disparity estimate. Such a confidence map represents valuable, additional information that can be leveraged in subsequent steps of the 3D processing chain. In this regard, this paper presents an extension of a cross bilateral filter that leverages this reliability information during a fast-converging refinement step in order to create robust and reliable disparity maps.

Reliability-aware cross multilateral filtering for robust disparity map refinement

As local correspondence methods alone often do not achieve a sufficient robustness and accuracy, a subsequent refinement stage is commonly employed to improve upon the initial disparity estimates. In recent years, cross bilateral filters have already been successfully applied for this purpose. They do not only smooth the disparity maps but also help to better align the disparity values with the object edges resulting in an improved object segmentation. In this paper, a refinement step based on an extension of a cross bilateral filter is proposed that utilizes explicit reliability information in order to create robust and reliable disparity maps.

A Unified Trifocal System for Advanced Depth-Based 3D Capture

While benefits of depth-maps for content creation and 3D post-production for multi-view displays are already widely known, their usage as an alternative stereo 3D (S3D) production workflow is just starting to become visible. In addition the generation of reliable depth-maps is still an active research topic. — The paper describes a comprehensive 3D capture infrastructure approach comprising a trifocal camera system entirely based on Ethernet interface technology, a dedicated field recording solution realized with solid state technology capable of recording up to seven uncompressed ‘high frame rate’ camera streams in parallel, and algorithms for trifocal depth estimation enabling real-time depth preview on-set as well as high-quality depth extraction offline. The new interface approach harmonizes and consolidates image data, depth information and general metadata. The system provides a new 3D production workflow with no need of scene dependent adjustments but the possibility to calculate a virtual view with any favored baseline during post.