Nanoscale lamellar structures are ridge like structures that have shown great promise for device oriented applications. In order to study multiple device oriented structures simultaneously, fingerprint-like lamellar structures are produced. The device oriented structural features that we are interested in are subsets of the fingerprint patterns. Manual detection and evaluation of critical device oriented structural features in the fingerprint pattern is impractical and hence we need automated image analysis techniques. In this thesis we develop automated image analysis techniques for the quantitative study of these lamellar structures. The contributions of this thesis are: an accurate propose-and-verify methodology for the extraction of lamellar structural features; a point-set registration technique that finds the (partial) overlap between two related lamellar patterns; and, the development of quantitative analysis and visualization tools for the lamellar structures. The structural features are accurately extracted using a combination of Gabor (proposal) and wedge (verification) filters. The registration is achieved by using a speeded-up, distance map based iterative closest point technique with uniform parameter space binning. Two application case studies, spatial variation of structural similarity and defect density estimation, are presented.
展开▼
