| |
| | Contour matching: problem and related work (Site not responding. Last check: 2007-10-23) |
 | | The idea behind proximity matching methods is to search for the best matching while permitting the rotation, translation and scaling (to be called alignment transformation) of each curve, such that the distances between matched key points are minimized [4,20,21,42]. |
 | | The vision problem is different from the string matching problem in two major aspects, however: first, in vision invariance to certain geometrical transformations is desired; second, a resolution degradation (or smoothing) may create a completely different list of elements in the syntactical representation. |
 | | If invariant attributes are used, the first problem is immediately addressed, but then the resolution problem either remains unsolved [1,17,25] or may be addressed by constructing for each curve a cascade of representations at different scales [3,29,41]. |
| www.dai.ed.ac.uk /CVonline/LOCAL_COPIES/GDALYAHU/gdalyahu.html (1104 words) |
|