| 摘要 |
PURPOSE:To decide the correspondence of feature points with high accuracy, the minimum sum of distances among feature points and no leakage, by performing rearrangement operations of feature points in the order of smaller values in both row and column directions of an inter-feature point distance matrix as well as comparing operations among inter-feature point distances by the finite number of times respectively. CONSTITUTION:An inter-feature point distance matrix of M rows/N columns is calculated between a feature point gathering I containing M feature points and a feature point gathering II consisting of N feature points (M>=N) before and after the displacement/deformation of a pattern. Thus the minimum element of each row is detected so that the initial correspondence candidate point is decided within the gathering II against each feature point of gathering I. Then short correspondence solving processing is repeated to secure the correspondence between the feature points of gathering II which never emerge in the initial correspondence and each feature point of the minimum distance in the gathering I. At the same time, the correspondence candidate points are selected out of the gathering I in the order of small distances for each feature point of the gathering II and the smaller value is left with preference between two overlapping candidate points if detected. Overlap correspondence solving processing is applied to correct the initial correspondence when the overlapping correspondence is solved. Then the correspondence result having smaller inter-feature point distance sum is outputted as the feature point correspondence result out of two types of correspondence results.
|
