Multiple View Geometry in Computer Vision (2nd ed)

By Richard I. Hartley, Andrew Zisserman. 2003

2,485 citations recorded by the source corpus.

View Paper: https://www.semanticscholar.org/paper/4fc2074e7f057ae255a6a79efecdf5df3d121895

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

On-Manifold Preintegration for Real-Time Visual--Inertial Odometry (2015, methodology): “This approach is well known in the bundle adjustment literature as the Schur complement trick, where a standard practice is to update the linearization point of ρl via back-substitution [61].”
Particle Filter SLAM with High Dimensional Vehicle Model (2009, methodology): “This allows us to model the camera using the pinhole model with unit focal length, according to [25] and [21],”
Pixel2Mesh++: Multi-View 3D Mesh Generation via Deformation (2019, background): “Traditional multi-view stereo (MVS) [14] relies on correspondences built via photo-consistency and thus it is vulnerable to large baselines, occlusions, and texture-less regions.”
Event-Based Vision: A Survey (2019, methodology): “These works follow the classical two-step stereo solution: ﬁrst solve the event correspondence problem across image planes (i.e., epipolar matching) and then triangulate the location of the 3D point [176].”
Unsupervised Deep Homography: A Fast and Robust Homography Estimation Model (2017, methodology): “The DLT algorithm is used to solve for the homography matrix H given a set of four point correspondences [30].”
Estimation of Tree Stem Attributes Using Terrestrial Photogrammetry with a Camera Rig (2016, unknown): “–24–1) were detected automatically using adaptive RANSAC on the rigid-body transformation [23,25] (p = 99.”
Geometric context from a single image (2005, background): “Although the 3D orientation of a plane (relative to the viewer) can be completely determined by its vanishing line [10], such information cannot easily be extracted from relatively unstructured outdoor images.”
Automatic photo pop-up (2005, background): “…the most accurate (but also the most labor-intensive) approach, recovering a metric reconstruction of an architectural scene by using projective geometry constraints [Hartley and Zisserman 2004] to compute 3D locations of user-specified points given their projected distances from the ground plane.”
Kimera: an Open-Source Library for Real-Time Metric-Semantic Localization and Mapping (2019, methodology): “At each iSAM2 iteration, the structureless vision model estimates the 3D position of the observed features using DLT [58] and analytically eliminates the corresponding 3D points from the VIO state [59].”
Recovering Surface Layout from an Image (2007, methodology): “Knowledge of the vanishing line of a plane completely specifies its 3D orientation relative to the viewer (Hartley and Zisserman, 2004), but such information cannot easily be extracted from outdoor, relatively unstructured images.”
Segmentation-Based Image Copy-Move Forgery Detection Scheme (2015, methodology): “This process follows the traditional way of computer vision [21].”
Visual Odometry : Part II: Matching, Robustness, Optimization, and Applications (2012, methodology): “Inlier points to a hypothesis are found by computing the point-to-epipolar line distance [35].”
Segmentation and Reconstruction of Polyhedral Building Roofs From Aerial Lidar Point Clouds (2010, methodology): “The RANSAC method needs to be applied to the entire data set and generates only one set of parameters when the process is finished [20].”
Go-ICP: A Globally Optimal Solution to 3D ICP Point-Set Registration (2016, methodology): “If no rough guess is available, one can gradually increase r until a measure such as the inlier number or RMS error attains a set value, or apply the automatic overlap estimation proposed in [62].”
Toward automated driving in cities using close-to-market sensors: An overview of the V-Charge Project (2013, methodology): “of car poses are improved with a full Bundle Adjustment [13] where the total reprojection errors are minimized.”