Herbert Bay's research while affiliated with ETH Zurich and other places

Publications (14)

Patent
A method for operating on images is described for interest point detection and/or description working under different scales and with different rotations, e.g. for scale-invariant and rotation-invariant interest point detection and/or description.
Article
This article presents a novel scale- and rotation-invariant detector and descriptor, coined SURF (Speeded-Up Robust Features). SURF approximates or even outperforms previously proposed schemes with respect to repeatability, distinctiveness, and robustness, yet can be computed and compared much faster. This is achieved by relying on integral images...
Conference Paper
We present a system which allows to request information on physical objects by taking a picture of them. This way, using a mobile phone with integrated camera, users can interact with objects or "things" in a very simple manner. A further advantage is that the objects themselves don't have to be tagged with any kind of markers. At the core of our s...
Conference Paper
Dense depth maps can be estimated in a Bayesian sense from multiple calibrated still images of a rigid scene relative to a reference view [1]. This well-established probabilistic framework is extended by adaptively refining a triangular meshing procedure and by automatic cross-validation of model parameters. The adaptive refinement strategy locally...
Conference Paper
Full-text available
In this paper, we present a novel scale- and rotation-invariant interest point detector and descriptor, coined SURF (Speeded Up Robust Features). It approximates or even outperforms previously proposed schemes with respect to repeatability, distinctiveness, and robustness, yet can be computed and compared much faster. This is achieved by relying o...
Conference Paper
Full-text available
This paper addresses the problem of camera self-calibration, bundle adjustment and 3D reconstruction from line segments in two images of poorly-textured indoor scenes. First, we generate line segment correspondences, using an extended version of our previously proposed matching scheme. The first main contribution is a new method to identify polyhed...
Conference Paper
Full-text available
Laser photocoagulation is a proven procedure to treat various pathologies of the retina. Challenges such as motion compensation, correct energy dosage, and avoiding incidental damage are responsible for the still low success rate. They can be overcome with improved instrumentation, such as a fully automatic laser photocoagulation system. In this pa...
Conference Paper
This article presents a novel scale- and rotation-invariant detector and descriptor, coined SURF (Speeded-Up Robust Features). SURF approximates or even outperforms previously proposed schemes with respect to repeatability, distinctiveness, and robustness, yet can be computed and compared much faster.This is achieved by relying on integral images f...
Conference Paper
Full-text available
We present a new method for matching line segments be- tween two uncalibrated wide-baseline images. Most cur- rent techniques for wide-baseline matching are based on viewpoint invariant regions. Those methods work well with highly textured scenes, but fail with poorly textured ones. We show that such scenes can be successfully matched us- ing line...
Article
Full-text available
In this paper, we describe the application of the novel SURF (Speeded Up Robust Features) algorithm [1] for the recognition of objects of art. For this purpose, we developed a prototype of a mobile interactive museum guide consisting of a tablet PC that features a touchscreen and a webcam. This guide recognises objects in museums based on images ta...
Article
Full-text available
In this paper, we describe the prototype of an interactive museum guide. It runs on a tablet PC that features a touchscreen, a webcam and a Bluetooth receiver. This guide recognises objects on display in museums based on images of the latter which are taken directly by the visitor. Furthermore, the computer can determine the visitor's lo-cation by...

Citations

... Traditionally, the localization problem has been tackled using 3D geometry [44], [172]. Common methods are Structure from Motion (SfM) [159], [173], [195] and Simultaneously Localization and Mapping (SLAM) [51], [88], [131]- [133], [139], [146], [194] that are based on image matching techniques as Scale-invariant Feature Transform (SIFT) [192], Speed Up Robust Feature (SURF) [5] and Oriented FAST, Rotated BRIEF (ORB) [24], [131], [132]. SLAM-driven 3D point registration methods [95] enable precise self-localization even in unknown environments. ...
... Interest points (Funayama et al., 2012) need to have an exact location in the image, which is extracted from the response maps by finding all the local maxima. SURF finds local maxima using a non-maximum suppression in a 3 × 3 × 3 neighbourhood around each pixel (Arnesen, 2010). ...
... Commonly used remote sensing image matching methods can be divided into two categories: one is feature-based, and the other is region-based [10]. In the field of computer vision, classic feature matching algorithms such as SIFT (Scale-Invariant Feature Transform) [11], SURF (Speeded Up Robust Features) [12], FAST (Features from Accelerated Segment Test), and ORB (Oriented fast and Rotated BRIEF) [13,14] are widely used in the field of remote sensing image registration, while these descriptors are very sensitive to the radiation difference between images. In order to overcome the influence of radiation differences between multimodal images, many scholars have made improvements on the SIFT algorithm [15][16][17]. ...
... researchers devote most of the time to the trade-off between accuracy and computational cost. Scale Invariant Feature Transform (Lowe, 2004) and Speeded-Up Robust Features (Bay et al., 2006) are most famous algorithms that are used for feature detection in handcraft designing. Expert selects the detected features for extracting the characteristics of data involved. ...
... Automated registration methods available for UAV multispectral image registration may improve geometric consistency and be more time-effective (Angel et al. 2020; Meng et al. 2021;Padró et al. 2019). Such approaches include the scale-invariant feature transform (SIFT) (Lowe 1999), speeded up robust features (SURF) (Bay et al. 2008), multi-scale SIFT-RANSACT (RANdom SAmple Consensus) methods (Oh, Toth, and Grejner-Brzezinska 2011), oriented fast and rotated brief (ORB) (Rublee et al. 2011), and channel features of orientated gradients (CFOG) (Ye et al. 2019). Approaches like these may hold promise for effective and accurate geo-referencing of UAV and satellite image data. ...
... Previously, some mobile museum guide systems have been introduced to help visitors [14,15,16,17]. ...
... U-SURF is a simplified version of SURF, which does not consider image rotation, so it can improve efficiency. The PD PRPS patterns collected in the substation and in the laboratory are all upright images, and there is no change in the viewing angle; therefore, the simpler U-SURF feature extraction method was selected to simplify the experimental process and increase the speed [14,15]. This research applied the local feature extraction method of images to the field of PD pattern recognition for the first time. ...
... The feature point describes a location where its adjacent area has unique intensity changes that are invariant against image translation, rotation, and scaling; hence the same feature point can be consistently found and paired in different video frames. In addition to Shi-Tomasi features, other common 2D image features include features from accelerated segment test (FAST) [66], Harris-Stephens [67], binary robust invariant scalable keypoints (BRISK) [68] and speeded up robust features (SURF) [69]. The comparison of these features in image matching in vision-based SHM problems is investigated in [44]. ...
... In this paper, we are aiming at a disparity map that facilitates reconstruction and as such accurate disparity in non-occluded regions takes more preference over detection of occlusion. It has been reported in [11, 12] that triangulation schemes may be used to accelerate multiview reconstruction algorithms such as those proposed in [13]. [12] adaptively subdivides the scene as it estimates disparity. ...
... It also does not prevent self-intersections nor guarantees watertightness. The authors of [BENV06] segment images into likely planar polygons based on 3D corner junctions and use best supporting lines to reconstruct polygons in 3D. For 2.5D reconstruction, extracted 3D lines [SZ97] are used with a dense height map to build a line arrangement on the ground plane and create geometric primitives and building masks [ZBKB08]. ...