Abstract
A critical comparison of methods of high-level interpretation of segmented images is presented and illustrated by examples drawn from two-dimensional image analysis. The methods discussed include boundary correlation, generalized Hough transformation, relational distance measures, graph matching, heuristic search and relaxation labelling. Each technique is considered in terms of its reliability in interpreting noisy and cluttered scenes and in terms of the dynamic complexity of the algorithm as a function of the input data. Although 2-D analysis of linearly segmented scenes is used as a mechanism for comparison, many of the arguments are applicable to three-dimensional image interpretation based on surface modelling for example. © 1988.
| Original language | English |
|---|---|
| Pages (from-to) | 241-259 |
| Number of pages | 19 |
| Journal | Pattern Recognition |
| Volume | 21 |
| Issue number | 3 |
| Publication status | Published - 1988 |
Keywords
- Computer vision
- Image interpretation
- Knowledge-based systems