Download presentation
Presentation is loading. Please wait.
1
Point Pattern Matching by Using Parameterization
The conference on Korea Multimedia Society(KMMS) 24~25 November 2006, Hansugn UNIV., Seoul, Korea Graduate School of Design & IT, Dongseo UNIV., Busan, Korea Joon-Jae Lee, Byung-Gook Lee Nam-Seok Choi
2
What is Fingerprint? Original Image Segmentation Binarization Thinning
3
Point Pattern of Fingerprints
4
Point Pattern Matching
We want to find a similarity transformation Ts,θ,tx,ty, such that T(P) matches Q. s is a scaling factor θ a rotation angle tx and ty the x and y transitions For ,
5
A fast expected time algorithm
6
Delaunay Triangulation
Voronoi diagrams 내의 3개의 사이트의 외접원 안에 다른 사이트가 존재하지 않는다면, 그 다이어그램 내의 사이트들을 인접한 영역의 사이트 들과 직선으로 연결함으로 삼각영역을 형성한다.
7
Delaunay Triangulation
Improving Worst-Case Optimal Delaunay Triangulation Algorithms (Geoff Leach) divide-and-conquer partition two halves remove and new L-R cross edges add
8
Parameterization Parameterization of mesh:
Set of distinct points xi in R3 Create triangulation T of the point set. Determine the corresponding points U on R2
9
Unit circle mapping
10
Neighborhoods mapping result
Rotation factor = radian
Similar presentations