Automatic registration of aerial photographs and digitized maps

Markov Random Field Modeling in Computer Vision

“…Li makes use of contextual constraints not only on the prior configuration of labelings but also on the observed data into the labeling process (Li 1992a;Li 1992b;Li 1992c;Li et al 1993). He proposes an energy function, on a heuristic basis, which combines contextual constraints from both the prior knowledge and the observation.…”

Multidimensional Co-occurrence Matrices for Object Recognition and Matching

“…Among characteristics so that the classes of objects are linearly the different representations proposed in the literature, separable. In other words, we argue that careful feature some of the most popular ones are attribute relational selection followed by a simple classifier is a much more databases (e.g., [3]) and graphs (e.g., [5]), semantic nets preferable approach than a quick feature selection stage (e.g., [6]), schemata (e.g., [7]), frames (e.g., [8]), Freeman followed by a carefully designed classifier. As is well chain codes ( [9]), and co-occurrence matrices.…”

Scene reconstruction from partially overlapping images with use of composite filters

“…Most production registration systems are semimanual in that a human operator selects the control points from which the computer registers the imagery. Research into automating this area includes automatic control point selection, 3 feature encoding for search-space reduction, 4 and efficient search strategies such as relaxation 5 and clustering 6 for information stored in image graphs 4,5,7 and interpretation trees. 8 Most of these approaches require a priori information about the template distortions or the specific features within them.…”