function [phi,S] = vl_tps(x1,x2,Y) % VL_TPS Compute the thin-plate spline basis % PHI=VL_TPS(X1,X2,Y) returns the basis PHI of a thin-plate spline % (TPS) defined on the domain X1,X2 with control points Y. % % X1 and X2 are MxN matrices specifying the grid vertices. When % warping images, these usually correspond to image pixels. % % Y is a 2xK matrix specifying the control points, one per % column. Ofthen Y is a subset of the domain X1,X2, but this is not % required. % % PHI is a (K+3)xNxM matrix, with one layer per basis element. Each % basis element is a function of the domain X1,X2. % % [PHI,S] = VL_TPS(X1,X2,Y) additionally returns the stiffness matrix S % of the TPS. % % See also: VL_WTPS(), VL_HELP(). % Copyright (C) 2007-12 Andrea Vedaldi and Brian Fulkerson. % All rights reserved. % % This file is part of the VLFeat library and is made available under % the terms of the BSD license (see the COPYING file). X = [x1(:)';x2(:)'] ; K = size(Y,2) ; Q = size(X,2) ; U = vl_tpsu(Y,Y) ; L = [[ones(1,K); Y], zeros(3) ; U, ones(K,1), Y'] ; invL = inv(L) ; tmp = vl_tpsu(Y,X) ; phi = invL * [ ones(1,Q) ; X(1,:) ; X(2,:) ; tmp ] ; [M,N] = size(x1) ; phi = reshape(phi,K+3,M,N) ; if nargout > 1 % See Bookstein; note that here the terms are re-arranged a bit invLn = invL(1:K, end-K+1:end) ; S = invLn * U * invLn ; end