Approximate Correspondences in High Dimensions
نویسندگان
چکیده
Pyramid intersection is an efficient method for computing an approximate partial matching between two sets of feature vectors. We introduce a novel pyramid embedding based on a hierarchy of non-uniformly shaped bins that takes advantage of the underlying structure of the feature space and remains accurate even for sets with high-dimensional feature vectors. The matching similarity is computed in linear time and forms a Mercer kernel. We also show how the matching itself (a correspondence field) may be extracted for a small increase in computational cost. Whereas previous matching approximation algorithms suffer from distortion factors that increase linearly with the feature dimension, we demonstrate that our approach can maintain constant accuracy even as the feature dimension increases. When used as a kernel in a discriminative classifier, our approach achieves improved object recognition results over a state-of-the-art set kernel.
منابع مشابه
Investigating the Role of Correspondences in Iranian Political Relations with Neighboring Governments in the First Shah Abbas Era (with an Emphasis on the Manuscripts of the Correspondences)
The kingship of Shah Abbas was contemporary with one of the most critical period in the Safavid era. One of the main challenges of the Safavid government at this time was how to deal with the neighboring states. One of Shah Abbas's first steps to address the challenges was to enable diplomacy and to send numerous communications to the neighboring states. In addition to introducing the correspon...
متن کاملThe size-brightness correspondence: evidence for crosstalk among aligned conceptual feature dimensions.
The same core set of cross-sensory correspondences connecting stimulus features across different sensory channels are observed, regardless of the modality of the stimulus with which the correspondences are probed. This observation suggests that correspondences involve modality-independent representations of aligned conceptual feature dimensions and predicts a size-brightness correspondence in w...
متن کاملThe colors and textures of musical sounds
Music-to-color associations show emotionally-mediated cross-modal correspondences (Palmer et al., 2013): people choose colors as going best with music when their emotional content matches (e.g., happy-looking colors go best with happysounding music). What musical/acoustic features underlie such correspondences? And are music-to-texture correspondences also evident? Experiments using highly-cont...
متن کاملApproximate K-D Tree Search for Efficient ICP
A method is presented that uses an Approximate Nearest Neighbor method for determining correspondences within the Iterative Closest Point Algorithm. The method is based upon the k-d tree. The standard k-d tree uses a tentative backtracking search to identify nearest neighbors. In contrast, the Approximate k-d tree (Ak-d tree) applies a depthfirst nontentative search to the k-d tree structure. T...
متن کامل