This paper proposes a method for line matchingbased on invariance of feature points. Firstly, feature pointsare roughly matched by Normalized Cross Correlation(NNC) and Average of Square Difference (ASD).Additionally, feature points obtained from the two views aregrouped into matched point pairs. Finally, curve segmentsbetween matched point pairs are matched by dynamicprogramming algorithm with edge potential functions (EPF)taken as the measure. The proposed method makes full useof feature points, the relationship between feature pointsand the curve, and space information of the gray image.
展开▼