| Autor: |
ARICAN, Erkut, OZDEN, Kemal Egemen |
| Předmět: |
|
| Zdroj: |
International Symposium on Computing in Science & Engineering; Oct2013, p41-46, 6p |
| Abstrakt: |
This paper studies a computer vision based Augmented Reality system. Rather than using a single anchor image, a semi-complete panoramic representation created from multiple images is used. These images are typically generated by a rotation and zooming camera, which is enough to many tourism and GIS applications while keeping the image deformation geometry simple. Panoramic images as anchor is getting more attention in AR field and novel aspects of our work can be described as follows: 1- Multi-scale panoramic representation is tree structured and consists of purely local image features 2- Tree levels directly relate to zoom levels of the images, hence an automatic relative zoom level detection based on mean-shift algorithm is introduced. 3- The "Augmented Knowledge" is multi-scale. The system does not suffer from high error rates for close-up scenes in comparison to classical non-vision techniques. Image lookup has logarithmic complexity rather than linear thanks to inherent tree structure. Since only local images features are kept, many complicated steps for actual panorama creation are avoided. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
