The combinatorics of object recognition in cluttered environments using constrained search
Artificial Intelligence
Similarity, connectionism, and the problem of representation in vision
Neural Computation
Maintaining knowledge about temporal intervals
Communications of the ACM
Combining Contour and Region Information for Perceptual Grouping
Mustererkennung 1998, 20. DAGM-Symposium
Hi-index | 0.00 |
In this paper, we address the visual analogies recognition problem. Often morphogeometric-based metrics are not sufficient to express high-level relations as well as complex queries. The model we are proposing extends the original virtual image formalism, into a hierarchical list of attributes. The lowest level is composed of spatial relations, whereas the highest level model relations between groups in terms of homogeneity and cardinality. The information stored in each layer is the basis for the new similarity metric, that models the query by analogy paradigm (QBA for short) we present in this work. This new analogy-based index, allows users to express complex relations as well as search for pictures expressing similar relations, such as functional association or group membership between objects.