A new hierarchical approach in robust real-time image feature detection and matching

Object recognition forms a ubiquitous problem in digital image processing. The detection of robust image features of high distinctiveness is one important key in this regard. We present a new hierarchical approach in object recognition targeting at high robustness, yet trying to fulfill hard real-time constraints. The former will be achieved using SIFT and SURF operators, while the latter is done by employing a fast pre-processing step exploiting decision-trees.