1
subroutine regrid(iopt,mx,x,my,y,z,xb,xe,yb,ye,kx,ky,s,
2
* nxest,nyest,nx,tx,ny,ty,c,fp,wrk,lwrk,iwrk,kwrk,ier)
3
c given the set of values z(i,j) on the rectangular grid (x(i),y(j)),
4
c i=1,...,mx;j=1,...,my, subroutine regrid determines a smooth bivar-
5
c iate spline approximation s(x,y) of degrees kx and ky on the rect-
6
c angle xb <= x <= xe, yb <= y <= ye.
7
c if iopt = -1 regrid calculates the least-squares spline according
8
c to a given set of knots.
9
c if iopt >= 0 the total numbers nx and ny of these knots and their
10
c position tx(j),j=1,...,nx and ty(j),j=1,...,ny are chosen automatic-
11
c ally by the routine. the smoothness of s(x,y) is then achieved by
12
c minimalizing the discontinuity jumps in the derivatives of s(x,y)
13
c across the boundaries of the subpanels (tx(i),tx(i+1))*(ty(j),ty(j+1).
14
c the amounth of smoothness is determined by the condition that f(p) =
15
c sum ((z(i,j)-s(x(i),y(j))))**2) be <= s, with s a given non-negative
16
c constant, called the smoothing factor.
17
c the fit is given in the b-spline representation (b-spline coefficients
18
c c((ny-ky-1)*(i-1)+j),i=1,...,nx-kx-1;j=1,...,ny-ky-1) and can be eval-
19
c uated by means of subroutine bispev.
22
c call regrid(iopt,mx,x,my,y,z,xb,xe,yb,ye,kx,ky,s,nxest,nyest,
23
c * nx,tx,ny,ty,c,fp,wrk,lwrk,iwrk,kwrk,ier)
26
c iopt : integer flag. on entry iopt must specify whether a least-
27
c squares spline (iopt=-1) or a smoothing spline (iopt=0 or 1)
29
c if iopt=0 the routine will start with an initial set of knots
30
c tx(i)=xb,tx(i+kx+1)=xe,i=1,...,kx+1;ty(i)=yb,ty(i+ky+1)=ye,i=
31
c 1,...,ky+1. if iopt=1 the routine will continue with the set
32
c of knots found at the last call of the routine.
33
c attention: a call with iopt=1 must always be immediately pre-
34
c ceded by another call with iopt=1 or iopt=0 and
37
c mx : integer. on entry mx must specify the number of grid points
38
c along the x-axis. mx > kx . unchanged on exit.
39
c x : real array of dimension at least (mx). before entry, x(i)
40
c must be set to the x-co-ordinate of the i-th grid point
41
c along the x-axis, for i=1,2,...,mx. these values must be
42
c supplied in strictly ascending order. unchanged on exit.
43
c my : integer. on entry my must specify the number of grid points
44
c along the y-axis. my > ky . unchanged on exit.
45
c y : real array of dimension at least (my). before entry, y(j)
46
c must be set to the y-co-ordinate of the j-th grid point
47
c along the y-axis, for j=1,2,...,my. these values must be
48
c supplied in strictly ascending order. unchanged on exit.
49
c z : real array of dimension at least (mx*my).
50
c before entry, z(my*(i-1)+j) must be set to the data value at
51
c the grid point (x(i),y(j)) for i=1,...,mx and j=1,...,my.
53
c xb,xe : real values. on entry xb,xe,yb and ye must specify the bound-
54
c yb,ye aries of the rectangular approximation domain.
55
c xb<=x(i)<=xe,i=1,...,mx; yb<=y(j)<=ye,j=1,...,my.
57
c kx,ky : integer values. on entry kx and ky must specify the degrees
58
c of the spline. 1<=kx,ky<=5. it is recommended to use bicubic
59
c (kx=ky=3) splines. unchanged on exit.
60
c s : real. on entry (in case iopt>=0) s must specify the smoothing
61
c factor. s >=0. unchanged on exit.
62
c for advice on the choice of s see further comments
63
c nxest : integer. unchanged on exit.
64
c nyest : integer. unchanged on exit.
65
c on entry, nxest and nyest must specify an upper bound for the
66
c number of knots required in the x- and y-directions respect.
67
c these numbers will also determine the storage space needed by
68
c the routine. nxest >= 2*(kx+1), nyest >= 2*(ky+1).
69
c in most practical situation nxest = mx/2, nyest=my/2, will
70
c be sufficient. always large enough are nxest=mx+kx+1, nyest=
71
c my+ky+1, the number of knots needed for interpolation (s=0).
72
c see also further comments.
74
c unless ier=10 (in case iopt >=0), nx will contain the total
75
c number of knots with respect to the x-variable, of the spline
76
c approximation returned. if the computation mode iopt=1 is
77
c used, the value of nx should be left unchanged between sub-
79
c in case iopt=-1, the value of nx should be specified on entry
80
c tx : real array of dimension nmax.
81
c on succesful exit, this array will contain the knots of the
82
c spline with respect to the x-variable, i.e. the position of
83
c the interior knots tx(kx+2),...,tx(nx-kx-1) as well as the
84
c position of the additional knots tx(1)=...=tx(kx+1)=xb and
85
c tx(nx-kx)=...=tx(nx)=xe needed for the b-spline representat.
86
c if the computation mode iopt=1 is used, the values of tx(1),
87
c ...,tx(nx) should be left unchanged between subsequent calls.
88
c if the computation mode iopt=-1 is used, the values tx(kx+2),
89
c ...tx(nx-kx-1) must be supplied by the user, before entry.
90
c see also the restrictions (ier=10).
92
c unless ier=10 (in case iopt >=0), ny will contain the total
93
c number of knots with respect to the y-variable, of the spline
94
c approximation returned. if the computation mode iopt=1 is
95
c used, the value of ny should be left unchanged between sub-
97
c in case iopt=-1, the value of ny should be specified on entry
98
c ty : real array of dimension nmax.
99
c on succesful exit, this array will contain the knots of the
100
c spline with respect to the y-variable, i.e. the position of
101
c the interior knots ty(ky+2),...,ty(ny-ky-1) as well as the
102
c position of the additional knots ty(1)=...=ty(ky+1)=yb and
103
c ty(ny-ky)=...=ty(ny)=ye needed for the b-spline representat.
104
c if the computation mode iopt=1 is used, the values of ty(1),
105
c ...,ty(ny) should be left unchanged between subsequent calls.
106
c if the computation mode iopt=-1 is used, the values ty(ky+2),
107
c ...ty(ny-ky-1) must be supplied by the user, before entry.
108
c see also the restrictions (ier=10).
109
c c : real array of dimension at least (nxest-kx-1)*(nyest-ky-1).
110
c on succesful exit, c contains the coefficients of the spline
111
c approximation s(x,y)
112
c fp : real. unless ier=10, fp contains the sum of squared
113
c residuals of the spline approximation returned.
114
c wrk : real array of dimension (lwrk). used as workspace.
115
c if the computation mode iopt=1 is used the values of wrk(1),
116
c ...,wrk(4) should be left unchanged between subsequent calls.
117
c lwrk : integer. on entry lwrk must specify the actual dimension of
118
c the array wrk as declared in the calling (sub)program.
119
c lwrk must not be too small.
120
c lwrk >= 4+nxest*(my+2*kx+5)+nyest*(2*ky+5)+mx*(kx+1)+
122
c where u is the larger of my and nxest.
123
c iwrk : integer array of dimension (kwrk). used as workspace.
124
c if the computation mode iopt=1 is used the values of iwrk(1),
125
c ...,iwrk(3) should be left unchanged between subsequent calls
126
c kwrk : integer. on entry kwrk must specify the actual dimension of
127
c the array iwrk as declared in the calling (sub)program.
128
c kwrk >= 3+mx+my+nxest+nyest.
129
c ier : integer. unless the routine detects an error, ier contains a
130
c non-positive value on exit, i.e.
131
c ier=0 : normal return. the spline returned has a residual sum of
132
c squares fp such that abs(fp-s)/s <= tol with tol a relat-
133
c ive tolerance set to 0.001 by the program.
134
c ier=-1 : normal return. the spline returned is an interpolating
136
c ier=-2 : normal return. the spline returned is the least-squares
137
c polynomial of degrees kx and ky. in this extreme case fp
138
c gives the upper bound for the smoothing factor s.
139
c ier=1 : error. the required storage space exceeds the available
140
c storage space, as specified by the parameters nxest and
142
c probably causes : nxest or nyest too small. if these param-
143
c eters are already large, it may also indicate that s is
145
c the approximation returned is the least-squares spline
146
c according to the current set of knots. the parameter fp
147
c gives the corresponding sum of squared residuals (fp>s).
148
c ier=2 : error. a theoretically impossible result was found during
149
c the iteration proces for finding a smoothing spline with
150
c fp = s. probably causes : s too small.
151
c there is an approximation returned but the corresponding
152
c sum of squared residuals does not satisfy the condition
154
c ier=3 : error. the maximal number of iterations maxit (set to 20
155
c by the program) allowed for finding a smoothing spline
156
c with fp=s has been reached. probably causes : s too small
157
c there is an approximation returned but the corresponding
158
c sum of squared residuals does not satisfy the condition
160
c ier=10 : error. on entry, the input data are controlled on validity
161
c the following restrictions must be satisfied.
162
c -1<=iopt<=1, 1<=kx,ky<=5, mx>kx, my>ky, nxest>=2*kx+2,
163
c nyest>=2*ky+2, kwrk>=3+mx+my+nxest+nyest,
164
c lwrk >= 4+nxest*(my+2*kx+5)+nyest*(2*ky+5)+mx*(kx+1)+
165
c my*(ky+1) +max(my,nxest),
166
c xb<=x(i-1)<x(i)<=xe,i=2,..,mx,yb<=y(j-1)<y(j)<=ye,j=2,..,my
167
c if iopt=-1: 2*kx+2<=nx<=min(nxest,mx+kx+1)
168
c xb<tx(kx+2)<tx(kx+3)<...<tx(nx-kx-1)<xe
169
c 2*ky+2<=ny<=min(nyest,my+ky+1)
170
c yb<ty(ky+2)<ty(ky+3)<...<ty(ny-ky-1)<ye
171
c the schoenberg-whitney conditions, i.e. there must
172
c be subset of grid co-ordinates xx(p) and yy(q) such
173
c that tx(p) < xx(p) < tx(p+kx+1) ,p=1,...,nx-kx-1
174
c ty(q) < yy(q) < ty(q+ky+1) ,q=1,...,ny-ky-1
176
c if s=0 : nxest>=mx+kx+1, nyest>=my+ky+1
177
c if one of these conditions is found to be violated,control
178
c is immediately repassed to the calling program. in that
179
c case there is no approximation returned.
182
c regrid does not allow individual weighting of the data-values.
183
c so, if these were determined to widely different accuracies, then
184
c perhaps the general data set routine surfit should rather be used
185
c in spite of efficiency.
186
c by means of the parameter s, the user can control the tradeoff
187
c between closeness of fit and smoothness of fit of the approximation.
188
c if s is too large, the spline will be too smooth and signal will be
189
c lost ; if s is too small the spline will pick up too much noise. in
190
c the extreme cases the program will return an interpolating spline if
191
c s=0 and the least-squares polynomial (degrees kx,ky) if s is
192
c very large. between these extremes, a properly chosen s will result
193
c in a good compromise between closeness of fit and smoothness of fit.
194
c to decide whether an approximation, corresponding to a certain s is
195
c satisfactory the user is highly recommended to inspect the fits
197
c recommended values for s depend on the accuracy of the data values.
198
c if the user has an idea of the statistical errors on the data, he
199
c can also find a proper estimate for s. for, by assuming that, if he
200
c specifies the right s, regrid will return a spline s(x,y) which
201
c exactly reproduces the function underlying the data he can evaluate
202
c the sum((z(i,j)-s(x(i),y(j)))**2) to find a good estimate for this s
203
c for example, if he knows that the statistical errors on his z(i,j)-
204
c values is not greater than 0.1, he may expect that a good s should
205
c have a value not larger than mx*my*(0.1)**2.
206
c if nothing is known about the statistical error in z(i,j), s must
207
c be determined by trial and error, taking account of the comments
208
c above. the best is then to start with a very large value of s (to
209
c determine the least-squares polynomial and the corresponding upper
210
c bound fp0 for s) and then to progressively decrease the value of s
211
c ( say by a factor 10 in the beginning, i.e. s=fp0/10,fp0/100,...
212
c and more carefully as the approximation shows more detail) to
213
c obtain closer fits.
214
c to economize the search for a good s-value the program provides with
215
c different modes of computation. at the first call of the routine, or
216
c whenever he wants to restart with the initial set of knots the user
218
c if iopt=1 the program will continue with the set of knots found at
219
c the last call of the routine. this will save a lot of computation
220
c time if regrid is called repeatedly for different values of s.
221
c the number of knots of the spline returned and their location will
222
c depend on the value of s and on the complexity of the shape of the
223
c function underlying the data. if the computation mode iopt=1
224
c is used, the knots returned may also depend on the s-values at
225
c previous calls (if these were smaller). therefore, if after a number
226
c of trials with different s-values and iopt=1, the user can finally
227
c accept a fit as satisfactory, it may be worthwhile for him to call
228
c regrid once more with the selected value for s but now with iopt=0.
229
c indeed, regrid may then return an approximation of the same quality
230
c of fit but with fewer knots and therefore better if data reduction
231
c is also an important objective for the user.
232
c the number of knots may also depend on the upper bounds nxest and
233
c nyest. indeed, if at a certain stage in regrid the number of knots
234
c in one direction (say nx) has reached the value of its upper bound
235
c (nxest), then from that moment on all subsequent knots are added
236
c in the other (y) direction. this may indicate that the value of
237
c nxest is too small. on the other hand, it gives the user the option
238
c of limiting the number of knots the routine locates in any direction
239
c for example, by setting nxest=2*kx+2 (the lowest allowable value for
240
c nxest), the user can indicate that he wants an approximation which
241
c is a simple polynomial of degree kx in the variable x.
243
c other subroutines required:
244
c fpback,fpbspl,fpregr,fpdisc,fpgivs,fpgrre,fprati,fprota,fpchec,
248
c dierckx p. : a fast algorithm for smoothing data on a rectangular
249
c grid while using spline functions, siam j.numer.anal.
250
c 19 (1982) 1286-1304.
251
c dierckx p. : a fast algorithm for smoothing data on a rectangular
252
c grid while using spline functions, report tw53, dept.
253
c computer science,k.u.leuven, 1980.
254
c dierckx p. : curve and surface fitting with splines, monographs on
255
c numerical analysis, oxford university press, 1993.
259
c dept. computer science, k.u. leuven
260
c celestijnenlaan 200a, b-3001 heverlee, belgium.
261
c e-mail : Paul.Dierckx@cs.kuleuven.ac.be
263
c creation date : may 1979
264
c latest update : march 1989
267
c ..scalar arguments..
268
real*8 xb,xe,yb,ye,s,fp
269
integer iopt,mx,my,kx,ky,nxest,nyest,nx,ny,lwrk,kwrk,ier
270
c ..array arguments..
271
real*8 x(mx),y(my),z(mx*my),tx(nxest),ty(nyest),
272
* c((nxest-kx-1)*(nyest-ky-1)),wrk(lwrk)
276
integer i,j,jwrk,kndx,kndy,knrx,knry,kwest,kx1,kx2,ky1,ky2,
277
* lfpx,lfpy,lwest,lww,maxit,nc,nminx,nminy,mz
278
c ..function references..
280
c ..subroutine references..
283
c we set up the parameters tol and maxit.
286
c before starting computations a data check is made. if the input data
287
c are invalid, control is immediately repassed to the calling program.
289
if(kx.le.0 .or. kx.gt.5) go to 70
292
if(ky.le.0 .or. ky.gt.5) go to 70
295
if(iopt.lt.(-1) .or. iopt.gt.1) go to 70
297
if(mx.lt.kx1 .or. nxest.lt.nminx) go to 70
299
if(my.lt.ky1 .or. nyest.lt.nminy) go to 70
301
nc = (nxest-kx1)*(nyest-ky1)
302
lwest = 4+nxest*(my+2*kx2+1)+nyest*(2*ky2+1)+mx*kx1+
303
* my*ky1+max0(nxest,my)
304
kwest = 3+mx+my+nxest+nyest
305
if(lwrk.lt.lwest .or. kwrk.lt.kwest) go to 70
306
if(xb.gt.x(1) .or. xe.lt.x(mx)) go to 70
308
if(x(i-1).ge.x(i)) go to 70
310
if(yb.gt.y(1) .or. ye.lt.y(my)) go to 70
312
if(y(i-1).ge.y(i)) go to 70
314
if(iopt.ge.0) go to 50
315
if(nx.lt.nminx .or. nx.gt.nxest) go to 70
322
call fpchec(x,mx,tx,nx,kx,ier)
323
if(ier.ne.0) go to 70
324
if(ny.lt.nminy .or. ny.gt.nyest) go to 70
331
call fpchec(y,my,ty,ny,ky,ier)
332
if (ier.eq.0) go to 60
334
50 if(s.lt.0.) go to 70
335
if(s.eq.0. .and. (nxest.lt.(mx+kx1) .or. nyest.lt.(my+ky1)) )
338
c we partition the working space and determine the spline approximation
342
jwrk = lwrk-4-nxest-nyest
347
call fpregr(iopt,x,mx,y,my,z,mz,xb,xe,yb,ye,kx,ky,s,nxest,nyest,
348
* tol,maxit,nc,nx,tx,ny,ty,c,fp,wrk(1),wrk(2),wrk(3),wrk(4),
349
* wrk(lfpx),wrk(lfpy),iwrk(1),iwrk(2),iwrk(3),iwrk(knrx),
350
* iwrk(knry),iwrk(kndx),iwrk(kndy),wrk(lww),jwrk,ier)