This thesis addresses the stereo matching problem, in which the depth information for scene objects is obtained by matching corresponding pixels in a pair of 2D images that were taken from slightly different viewpoints. In general, there are two branches of stereo matching algorithms, i.e., global and local matching methods. In this thesis we exclusively focus on local stereo matching methods, which rely on finding corresponding image pixels between the left and right stereo partner by calculating aggregated color dissimilarities of pixels within patches. While local stereo methods are attractive due to their computational speed, their main problem is to find a suitable patch size and shape to produce high-quality results even in challenging image regions such as object borders and homogeneous areas. The solution to overcome this problem is to use the principle of adaptive support weight aggregation. In this thesis, we propose novel adaptive support weight aggregation techniques that advance the state-of-the-art in both reconstruction quality and computational speed.
The contributions of this thesis can be summarized as follows. (i) We propose an improved method for computing the adaptive support weights by imposing an additional connectivity constraint that relies on a geodesic distance transform. The connectivity property results in more accurate support weights and, consequently, improved depth maps. (ii) In order to reduce the processing time, we present a method for considerably speeding up the computation time of an adaptive support weight algorithm by employing the adaptive weights for efficient color segmentation and then applying a segmentation-based sliding window technique for aggregating the costs. (iii) From the cost filtering/smoothing point of view, we introduce a stereo matching algorithm which uses the fast guided image filtering strategy for edge-preserving cost aggregation. This algorithm combines the two important properties of a stereo matcher, i.e., high quality of the matching results and real-time performance. (iv) We also propose an efficient algorithm for computing temporally consistent depth maps from video sequences. The main idea is to transfer the concept of the efficient cost filtering technique to the spatio-temporal domain. (v) To help researchers/developers selecting among different weight computation techniques, we conduct an extensive experimental evaluation of a variety of support weight computation methods suggested in the literature or proposed by ourselves. This evaluation examines the accuracy of the retrieved depth maps and the computational efficiency of a corresponding GPU-based implementation. (vi) Finally, we demonstrate that our proposed stereo algorithms that rely on cost filtering can be extended successfully to the closely related optical flow (motion) problem.
A. Hosni: "Novel Methods for Discontinuity Preserving Support Aggregation in Local Stereo Matching"; Supervisor, Reviewer: M. Gelautz, D. Scharstein, A. Uhl; Institut für Softwaretechnik und interaktive Systeme, 2013; oral examination: 04-18-2013.
Click into the text area and press Ctrl+A/Ctrl+C or ⌘+A/⌘+C to copy the BibTeX into your clipboard… or download the BibTeX.