Principal Warps: Thin-Plate Splines and the Decomposition of Deformations 김진욱 ( 이동통신망연구실 ; 박천현 (3D 모델링 및 처리연구실 ; 이승재 ( 멀티미디어 이동통신연구실 ; 최종윤 (3D 모델링 및 처리연구실 ;
Thin-Plate Splines Warping Contents Introduction Interpolation and special function U(r) Algebraic properties of TPS Mapping function f(x, y) Spectral analysis of principal warps Examples Conclusion
Thin-Plate Splines Warping Related Works Thin-Plate Spline Approximation for Image Registration [2] Uses approximation approach (this paper uses interpolation) Landmark-Based Elastic Registration Using Approximating This-Plate Splines [3] Refinement of [2] Consistent Landmark and Intensity-Based Image Registration [4] Uses Intensity based information additionally
Thin-Plate Splines Warping Thin-Plate Spline The name “thin plate spline” refers to a physical analogy involving the bending of a thin sheet of metal In physical setting, the deflection is in the direction orthogonal to the plane In coordinate transformation, one interprets the lifting of the plates as a displacement of the coordinate within the plane
Thin-Plate Splines Warping Thin-plate spline as an interpolant TPS is 2D analog of the cubic spline of 1D Purpose : describe the deformation specified by Finitely many point-correspondence With irregular spacing In this paper, we use special function U(r) Define U(0) = 0
Thin-Plate Splines Warping Special function U(r) Graph of –U(r) X : (0, 0, 0) Circle Zeros of function Radius : 1/√e
Thin-Plate Splines Warping Special function U(r) : example D k s are the corners (1, 0), (0, 1), (-1, 0), (0, -1) of the square z(x, y) approximates
Thin-Plate Splines Warping Special function U(r) Given a set of data points weighted combination of U(r) centered at each data points, Gives interpolation function that passes through the points exactly And linear combination of U(r)s minimize the “bending energy”, defined by
Thin-Plate Splines Warping Algebra of the Thin-Plate Spline for Arbitrary Sets of Landmarks Define matrices (K,P,L) P 1 =(x 1,y 1 ), …, P n =(x n,y n ) are n landmark points in the ordinary Euclidean plane r ij = |P i - P j | : the distance between points i and j
Thin-Plate Splines Warping Algebra of the Thin-Plate Spline for Arbitrary Sets of Landmarks Define the vector W = (w 1, …, w n ) and the coefficients a 1, a x, a y by the equation V = (v 1, …, v n ) is any n-vector Y = (V | ) T
Thin-Plate Splines Warping Algebra of the Thin-Plate Spline for Arbitrary Sets of Landmarks From the previous equation, (3) means interpolation property of TPS (4) means boundary condition of TPS
Thin-Plate Splines Warping The resulting function f(x,y) Define the function f(x,y) First 3 terms : describes global affine transform Rest terms : describes (nonglobal) non-linear transformation f(x i,y i ) = v i for all i (interpolation property) minimizes the bending energy, since U(r) is the fundamental solution of biharmonic equation
Thin-Plate Splines Warping The resulting function f(x,y) In the application we take the points (x i,y i ) to be landmarks and V to be the n X 2 matrix Each (x i ’, y i ’) is the landmark homologus to (x i, y i ) in another copy of R 2 Maps each point (x i, y i ) to its homolog(x i ’,y i ’) Least bent of all such mapping function
Thin-Plate Splines Warping Principal Warps The value of I f (bending energy) is Proportional to quadratic form Note that notation used represents vector as (v0, v1, v2, … ) L n is the upper left n by n submatrix of L inverse In non-degenerate cases, U’s can be represented as one dimensional displacement of any single landmark holding the others in fixed position
Thin-Plate Splines Warping Spectrum of bending energy matrix The vector is, (W | A) T Vector A describes the affine part of transformation Rotation, Translation, Scaling Vector W describes the non-linearity of transformation Matrix L n -1 KL n -1 means Bending energy as a function of changes in the coordinates of the landmarks 0-valued eigenvalues are corresponding to affine transform
Thin-Plate Splines Warping Spectrum of bending energy matrix Non-zero eigenvalues If there are N landmarks, We have N – 3 non-zero eigenvalues (and eigenvectors) Corresponding eigenvectors There are N components in eigenvectors, and they are coefficients for the N functions U based at N landmarks Called “principal warps” of the configuration of landmarks The transforms are affine-free (not-global transform) Higher eigenvalue means higher bending energy, and smaller physical scale of deformation
Thin-Plate Splines Warping Spectrum of bending energy matrix When the number of landmarks, `N’ is, N = 3 All transformations are described by affine transform No principal warps N = 4 Only one (trivial) principal warp N > 4 Successive eigenvectors describe the more smaller scale of deformation
Thin-Plate Splines Warping Examples Warped image Original image
Thin-Plate Splines Warping Examples The matrix L, vector V Positions of landmarks : V Matrix L
Thin-Plate Splines Warping Examples Global affine transform are vectors From the a of the vector (w|a) We get the affine part of the mapping function
Thin-Plate Splines Warping Examples SVD of linear parts in affine transform yields Rotation matrix (44.89º) Scaling matrix in x-coordinate in y-coordinate Rotation matrix (-53.34º)
Thin-Plate Splines Warping Examples Remaining terms The matrix has 5 eigenvalues 3 0-valued eigenvalues : for affine transform 2 non-zero eigenvalues which have their own eigenvectors f =(0.2152, , , , ) for f =( , , , , ) for
Thin-Plate Splines Warping Examples Net changes in x, y coordinate From the above net-change in each coordinate, we can calculate the weights of principal warps
Thin-Plate Splines Warping Examples Expand the function f x, f y of the mapping function in terms of principal warps Principal warps are used as basis of deformation description
Thin-Plate Splines Warping Real world examples APERT syndrome (trace of x – ray images)
Thin-Plate Splines Warping Real world examples Landmarks assignment
Thin-Plate Splines Warping Discussion Image Discrimination If landmarks can be chosen consistently on the left-hand form, the space of the decompositions explored in the paper is a natural context for interpretation of all multivariate findings
Thin-Plate Splines Warping Discussion Landmark Identification Some points biologically homologous between images are not clearly identifiable by local processing. When landmarks can be localized to edges, likely-hood ratio which can be computed by principal warp analysis can be used
Thin-Plate Splines Warping Discussion Description of Actual Deformation Can provide actual measure of deformation description, not an unwarp method
Thin-Plate Splines Warping Discussion Instantiation of Primitives In biomedical imaging, details are ordinaily strongly intercorrelated, and cannot observe all the details all at once. Regression analysis upon principal warps can help this problem
Thin-Plate Splines Warping Discussion 3 dimensions extension We used the 2-dimensional form r 2 log r 2 We can use U(r) = |r| in 3-dimension Called thin-`hyperplane’ spline
Thin-Plate Splines Warping Conclusion Method proposed in the paper, Decompose the deformation Introduces feature space for deformation Feature space is finite-dimensional space Easy to analyze Principal warps can represent the multivariate distribution of configuration of landmarks easily Landmark assignment problem For biologists, not for computer scientists