| 摘要 |
Matching combinations of regions are identified in a first and second image of a common scene. First and second sets of image regions are first identified in the first and second image on their own. Next a search is made for potentially matching first and second regions from the first and second set. First and second geometrical parameters such as moments are computed, which provide a first order approximation of the shapes of the first and second region respectively, in terms of the size and directions of major and minor axes of ellipses. These first and second geometrical parameters may be considered to define a first and second normalizing map, which map the first and second regions to isotropic regions, wherein the ellipses become circles. A search is made for closest third and fourth regions in the first and second image with matching distances to the first and second regions in spaces obtainable by mapping locations in the first and second image according to the first and second normalizing maps. The match is verified when such matching closest regions are found.
|
