Optimal Geometric Matching for Patch-Based Object Detection
Daniel Keysers, Thomas Deselaers, Thomas Michael Breuel · ELCVIA Electronic Letters on Computer Vision and Image Analysis · 2007
We present an efficient method to determine the optimal matching of two patch-based image object representations under rotation, scaling, and translation (RST). This use of patches is equivalent to a fullyconnected part-based model, for which the presented approach offers an efficient procedure to determine the best fit. While other approaches that use fully connected models have a high complexity in the number of parts used, we achieve linear complexity in that variable, because we only allow RST-matchings. The presented approach is used for object recognition in images: by matching images that contain certain objects to a test image, we can detect whether the test image contains an object of that class or not. We evaluate this approach on the Caltech data and obtain very competitive results.
Cited by 22 works. doi:10.5565/rev/elcvia.136 Open-access PDF
Authors
References
- Distinctive Image Features from Scale-Invariant Keypoints
- Random sample consensus
- Object class recognition by unsupervised scale-invariant learning
- An Algorithm for Vector Quantizer Design
- Unsupervised Learning of Models for Recognition
- Object Recognition with Features Inspired by Visual Cortex
- A survey of the hough transform
- Random Sample Consensus: A Paradigm for Model Fitting with Applications to Image Analysis and Automated Cartography
- A Sparse Object Category Model for Efficient Learning and Exhaustive Recognition
- Spatial Priors for Part-Based Recognition Using Statistical Models