Maximum Detector Response Markers for SURF |
In [1], we have presented a new class of visual markers aiming at maximum detectability by given feature extraction algorithms. Speeded Up Robust Features (SURF) are particularly interesting as they are very widely used for their high performance at comparably low complexity.
We have developed two basic markers adapted to SURF and designed them to trigger maximum detector response. More specifically, SURF examines the determinant of the Hessian at different scales and picks local maxima as features.
 | (1) |
Second derivatives of the Gaussian blurred images are approximated with box filters which, in combination with integral images, explains SURF's enormous speed-up in comparison to other algorithms operating in scale space.
Our SURF markers provably maximize the detector output (1) which makes them the ideal lightweight marker to introduce easily detectable features in otherwise unstructured parts of a scene.
 |
 |
| Fig. 1: The dark and light versions of our SURF marker. |
Fig. 2: Two marker arrays with detected SURF features. |
|
| [1] |
Florian Schweiger, Bernhard Zeisl, Pierre Georgel, Georg Schroth, Eckehard Steinbach, Nassir Navab |
|
Maximum Detector Response Markers for SIFT and SURF |
|
In Vision, Modeling and Visualization Workshop (VMV) |
|
November 2009 |
|
Braunschweig |
|
BibTeX PDF (© LMT) |
|
