Abstract
Abstract. We define a global matching framework based on energy pyramid, the Global Matching via Energy Pyramid (GM-EP) algorithm, which estimates the disparity map from a single stereo-pair by solving an energy minimization problem. We efficiently address this minimization by globally optimizing a coarse to fine sequence of sparse Conditional Random Fields (CRF) directly defined on the energy. This global discrete optimization approach guarantees that at each scale we obtain a near optimal solution, and we demonstrate its superiority over state of the art image pyramid approaches through application to real stereo-pairs. We conclude that multiscale approaches should be build on energy pyramids rather than on image pyramids.
Highlights
Precise Digital Surface Models (DSM) are widely employed in urban monitoring, geological surveys, architecture, or archeology
The matching energy of a pixel p is defined as: We use the vocabulary of the discrete optimization community and we introduce the notion of graph, unary term, edge cost, distance function, and label set for a Conditional Random Fields (CRF)
There are three main differences between the Global Matching via Energy Pyramid (GM-EP) and MicMac: (1) the GM-EP model is more complex than the one used in MicMac, i.e., both models seem equivalent if λ2 = 0 in Eq 7; (2), MicMac uses an image pyramid approach while the GM-EP works on an energy pyramid approach; and, (3) MicMac relies on semi-global optimization while GM-EP uses global optimization that produces near optimum solutions
Summary
Precise Digital Surface Models (DSM) are widely employed in urban monitoring, geological surveys, architecture, or archeology. The computer vision community has developed more advanced techniques to globally optimize the matching problem (Szeliski et al, 2008, Kappes et al, 2013) These techniques called global matching are mathematically more sound than semi-global matching as they guarantee a near global optimum (Boykov et al, 2001). Due to their computational complexity, they have only been applied to small images, i.e., less than 1000 × 1000 pixels, as proof of concept (Middleburry, 2014, Klaus et al, 2006, Kolmogorov and Zabih, 2001) and are not yet scalable to accommodate the large sizes of remote sensing images. We demonstrate the improvements of our approach through an application to real stereo acquisitions
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
