Scilab Website | Contribute with GitLab | Mailing list archives | ATOMS toolboxes
Scilab Online Help
2024.1.0 - Русский


splin2d

bicubic spline gridded 2d interpolation

Syntax

C = splin2d(x, y, z)
C = splin2d(x, y, z, spline_type)

Arguments

x, y

rows of nx and ny real numbers, in strictly increasing order: the x and y coordinates building the grid of the interpolation points (nodes).

z

a nx-by-ny matrix of real numbers: the function values.

spline_type

"not_a_knot" or "periodic" string: the type of spline to compute.

C

the coefficients of the bicubic patches. This output argument of splin2d is the input argument of the interp2d function.

Description

This function computes a bicubic spline or sub-spline s which interpolates the (x,y,zij ) points, ie, we have s(x,y)=zij  for all i=1,..,nx and j=1,..,ny. The resulting spline s is defined by the triplet (x,y,C) where C is the vector (of length 16(nx-1)(ny-1)) with the coefficients of each of the (nx-1)(ny-1) bicubic patches : on [x(i),x(i+1)] × [y(j),y(j+1)], s is defined by

s(x,y) = ∑_m=1→4 ∑_n=1→4 c_ij(m,n).(x-x_i)^{m-1}.(y-y_j)^{n-1}

The evaluation of s at some points must be done by the interp2d function. Several kind of splines may be computed by selecting the appropriate spline_type parameter. The method used to compute the bicubic spline (or sub-spline) is the old fashioned one's, i.e. to compute on each grid point (x,y) an approximation of the first derivatives ds/dx(x,y) and ds/dy(x,y) and of the cross derivative d2s/dxdy(x,y). Those derivatives are computed by the mean of 1d spline schemes leading to a C2 function (s is twice continuously differentiable) or by the mean of a local approximation scheme leading to a C1 function only. This scheme is selected with the spline_type parameter (see splin for details) :

"not_a_knot"

this is the default case.

"periodic"

to use if the underlying function is periodic : you must have z(1,j) = z(nx,j) for all j in [1,ny] and z(i,1) = z(i,ny) for i in [1,nx] but this is not verified by the interface.

Remarks

From an accuracy point of view use essentially the not_a_knot type or periodic type if the underlying interpolated function is periodic.

The natural, monotone, fast (or fast_periodic) type may be useful in some cases, for instance to limit oscillations (monotone being the most powerful for that).

To get the coefficients of the bi-cubic patches in a more friendly way you can use c = matrix(C, [4,4,nx-1,ny-1]) then the coefficient (k,l) of the patch (i,j) (see equation here before) is stored at c(k,l,i,j). Nevertheless the interp2d function wait for the big vector C and not for the hypermatrix c (note that one can easily retrieve C from c with C=c(:)).

Examples

// example 1 : interpolation of cos(x)cos(y)
n = 7;  // a regular grid with n x n interpolation points
        // will be used
x = linspace(0,2*%pi,n); y = x;
z = cos(x')*cos(y);
C = splin2d(x, y, z, "periodic");
m = 50; // discretization parameter of the evaluation grid
xx = linspace(0,2*%pi,m); yy = xx;
[XX,YY] = ndgrid(xx,yy);
zz = interp2d(XX,YY, x, y, C);
emax = max(abs(zz - cos(xx')*cos(yy)));
clf()
plot3d(xx, yy, zz, flag=[2 4 4])
[X,Y] = ndgrid(x,y);
param3d1(X,Y,list(z,-9*ones(1,n)), flag=[0 0])
str = msprintf(" with %d x %d interpolation points. ermax = %g",n,n,emax)
xtitle("spline interpolation of cos(x)cos(y)"+str)
// example 2 : different interpolation functions on random data
n = 6;
x = linspace(0,1,n); y = x;
z = rand(n,n);
np = 50;
xp = linspace(0,1,np); yp = xp;
[XP, YP] = ndgrid(xp,yp);
ZP1 = interp2d(XP, YP, x, y, splin2d(x, y, z, "not_a_knot"));
ZP2 = linear_interpn(XP, YP, x, y, z);
ZP3 = interp2d(XP, YP, x, y, splin2d(x, y, z, "natural"));
ZP4 = interp2d(XP, YP, x, y, splin2d(x, y, z, "monotone"));
gcf().color_map = jet(64);
clf()
subplot(2,2,1)
plot3d1(xp, yp, ZP1, flag=[2 2 4])
xtitle("not_a_knot")
subplot(2,2,2)
plot3d1(xp, yp, ZP2, flag=[2 2 4])
xtitle("bilinear interpolation")
subplot(2,2,3)
plot3d1(xp, yp, ZP3, flag=[2 2 4])
xtitle("natural")
subplot(2,2,4)
plot3d1(xp, yp, ZP4, flag=[2 2 4])
xtitle("monotone")
// example 3 : not_a_knot spline and monotone sub-spline
//             on a step function
a = 0; b = 1; c = 0.25; d = 0.75;
// create interpolation grid
n = 11;
x = linspace(a,b,n);
ind = find(c <= x & x <= d);
z = zeros(n,n); z(ind,ind) = 1;  // a step inside a square
// create evaluation grid
np = 220;
xp = linspace(a,b, np);
[XP, YP] = ndgrid(xp, xp);
zp1 = interp2d(XP, YP, x, x, splin2d(x,x,z));
zp2 = interp2d(XP, YP, x, x, splin2d(x,x,z,"monotone"));
// plot
clf()
gcf().color_map = jet(128);
subplot(1,2,1)
plot3d1(xp, xp, zp1, flag=[-2 6 4])
xtitle("spline (not_a_knot)")
subplot(1,2,2)
plot3d1(xp, xp, zp2, flag=[-2 6 4])
xtitle("subspline (monotone)")

See also

  • cshep2d — bidimensional cubic shepard (scattered) interpolation
  • linear_interpn — n dimensional linear interpolation
  • interp2d — bicubic spline (2d) evaluation function
Report an issue
<< splin Interpolation splin3d >>

Copyright (c) 2022-2024 (Dassault Systèmes)
Copyright (c) 2017-2022 (ESI Group)
Copyright (c) 2011-2017 (Scilab Enterprises)
Copyright (c) 1989-2012 (INRIA)
Copyright (c) 1989-2007 (ENPC)
with contributors
Last updated:
Mon Jun 17 17:55:06 CEST 2024