/*============================================================================= * * WCSLIB - an implementation of the FITS WCS proposal. * Copyright (C) 1995-1999, Mark Calabretta * * This library is free software; you can redistribute it and/or modify it * under the terms of the GNU Library General Public License as published * by the Free Software Foundation; either version 2 of the License, or (at * your option) any later version. * * This library is distributed in the hope that it will be useful, but * WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Library * General Public License for more details. * * You should have received a copy of the GNU Library General Public License * along with this library; if not, write to the Free Software Foundation, * Inc., 675 Mass Ave, Cambridge, MA 02139, USA. * * Correspondence concerning WCSLIB may be directed to: * Internet email: mcalabre@atnf.csiro.au * Postal address: Dr. Mark Calabretta, * Australia Telescope National Facility, * P.O. Box 76, * Epping, NSW, 2121, * AUSTRALIA * *============================================================================= * * C routines which implement the FITS World Coordinate System (WCS) * convention. * * Summary of routines * ------------------- * These utility routines apply the linear transformation defined by the WCS * FITS header cards. There are separate routines for the image-to-pixel, * linfwd(), and pixel-to-image, linrev(), transformations. * * An initialization routine, linset(), computes intermediate values from * the transformation parameters but need not be called explicitly - see the * explanation of lin.flag below. * * An auxiliary matrix inversion routine, matinv(), is included. It uses * LU-triangular factorization with scaled partial pivoting. * * * Initialization routine; linset() * -------------------------------- * Initializes members of a linprm data structure which hold intermediate * values. Note that this routine need not be called directly; it will be * invoked by linfwd() and linrev() if the "flag" structure member is * anything other than a predefined magic value. * * Given and/or returned: * lin linprm* Linear transformation parameters (see below). * * Function return value: * int Error status * 0: Success. * 1: Memory allocation error. * 2: PC matrix is singular. * * Forward transformation; linfwd() * -------------------------------- * Compute pixel coordinates from image coordinates. Note that where * celestial coordinate systems are concerned the image coordinates * correspond to (x,y) in the plane of projection, not celestial (lng,lat). * * Given: * imgcrd const double[] * Image (world) coordinate. * * Given and returned: * lin linprm* Linear transformation parameters (see below). * * Returned: * pixcrd d[] Pixel coordinate. * * Function return value: * int Error status * 0: Success. * 1: The transformation is not invertible. * * Reverse transformation; linrev() * -------------------------------- * Compute image coordinates from pixel coordinates. Note that where * celestial coordinate systems are concerned the image coordinates * correspond to (x,y) in the plane of projection, not celestial (lng,lat). * * Given: * pixcrd const double[] * Pixel coordinate. * * Given and/or returned: * lin linprm* Linear transformation parameters (see below). * * Returned: * imgcrd d[] Image (world) coordinate. * * Function return value: * int Error status * 0: Success. * 1: Error. * * Linear transformation parameters * -------------------------------- * The linprm struct consists of the following: * * int flag * This flag must be set to zero whenever any of the following members * are set or modified. This signals the initialization routine, * linset(), to recompute intermediaries. * int naxis * Number of image axes. * double *crpix * Pointer to the first element of an array of double containing the * coordinate reference pixel, CRPIXn. * double *pc * Pointer to the first element of the PC (pixel coordinate) * transformation matrix. * double *cdelt * Pointer to the first element of an array of double containing the * coordinate increments, CDELTn. * * The remaining members of the linprm struct are maintained by the * initialization routine and should not be modified. * * double *piximg * Pointer to the first element of the matrix containing the product * of the CDELTn diagonal matrix and the PC matrix. * double *imgpix * Pointer to the first element of the inverse of the piximg matrix. * * linset allocates storage for the above arrays using malloc(). Note, * however, that these routines do not free this storage so if a linprm * variable has itself been malloc'd then these structure members must be * explicitly freed before the linprm variable is free'd otherwise a memory * leak will result. * * Author: Mark Calabretta, Australia Telescope National Facility * $Id: lin.c,v 1.1.1.1 2003/07/09 19:06:26 bertin Exp $ *===========================================================================*/ #ifdef HAVE_CONFIG_H #include "config.h" #endif #ifdef HAVE_MATHIMF_H #include #else #include #endif #include #include "lin.h" /* Map error number to error message for each function. */ const char *linset_errmsg[] = { 0, "Memory allocation error", "PC matrix is singular"}; const char *linfwd_errmsg[] = { 0, "Memory allocation error", "PC matrix is singular"}; const char *linrev_errmsg[] = { 0, "Memory allocation error", "PC matrix is singular"}; int linset(lin) struct linprm *lin; { int i, ij, j, mem, n; n = lin->naxis; /* Allocate memory for internal arrays. */ mem = n * n * sizeof(double); lin->piximg = (double*)malloc(mem); if (lin->piximg == (double*)0) return 1; lin->imgpix = (double*)malloc(mem); if (lin->imgpix == (double*)0) { free(lin->piximg); return 1; } /* Compute the pixel-to-image transformation matrix. */ for (i = 0, ij = 0; i < n; i++) { for (j = 0; j < n; j++, ij++) { lin->piximg[ij] = lin->cdelt[i] * lin->pc[ij]; } } /* Compute the image-to-pixel transformation matrix. */ if (matinv(n, lin->piximg, lin->imgpix)) return 2; lin->flag = LINSET; return 0; } /*--------------------------------------------------------------------------*/ int linfwd(imgcrd, lin, pixcrd) const double imgcrd[]; struct linprm *lin; double pixcrd[]; { int i, ij, j, n; n = lin->naxis; if (lin->flag != LINSET) { if (linset(lin)) return 1; } for (i = 0, ij = 0; i < n; i++) { pixcrd[i] = 0.0; for (j = 0; j < n; j++, ij++) { pixcrd[i] += lin->imgpix[ij] * imgcrd[j]; } } for (j = 0; j < n; j++) { pixcrd[j] += lin->crpix[j]; } return 0; } /*--------------------------------------------------------------------------*/ int linrev(pixcrd, lin, imgcrd) const double pixcrd[]; struct linprm *lin; double imgcrd[]; { int i, ij, j, n; double temp; n = lin->naxis; if (lin->flag != LINSET) { if (linset(lin)) return 1; } for (i = 0; i < n; i++) { imgcrd[i] = 0.0; } for (j = 0; j < n; j++) { temp = pixcrd[j] - lin->crpix[j]; for (i = 0, ij = j; i < n; i++, ij+=n) { imgcrd[i] += lin->piximg[ij] * temp; } } return 0; } /*--------------------------------------------------------------------------*/ int matinv(n, mat, inv) const int n; const double mat[]; double inv[]; { register int i, ij, ik, j, k, kj, pj; int itemp, mem, *mxl, *lxm, pivot; double colmax, *lu, *rowmax, dtemp; /* Allocate memory for internal arrays. */ mem = n * sizeof(int); if ((mxl = (int*)malloc(mem)) == (int*)0) return 1; if ((lxm = (int*)malloc(mem)) == (int*)0) { free(mxl); return 1; } mem = n * sizeof(double); if ((rowmax = (double*)malloc(mem)) == (double*)0) { free(mxl); free(lxm); return 1; } mem *= n; if ((lu = (double*)malloc(mem)) == (double*)0) { free(mxl); free(lxm); free(rowmax); return 1; } /* Initialize arrays. */ for (i = 0, ij = 0; i < n; i++) { /* Vector which records row interchanges. */ mxl[i] = i; rowmax[i] = 0.0; for (j = 0; j < n; j++, ij++) { dtemp = fabs(mat[ij]); if (dtemp > rowmax[i]) rowmax[i] = dtemp; lu[ij] = mat[ij]; } /* A row of zeroes indicates a singular matrix. */ if (rowmax[i] == 0.0) { free(mxl); free(lxm); free(rowmax); free(lu); return 2; } } /* Form the LU triangular factorization using scaled partial pivoting. */ for (k = 0; k < n; k++) { /* Decide whether to pivot. */ colmax = fabs(lu[k*n+k]) / rowmax[k]; pivot = k; for (i = k+1; i < n; i++) { ik = i*n + k; dtemp = fabs(lu[ik]) / rowmax[i]; if (dtemp > colmax) { colmax = dtemp; pivot = i; } } if (pivot > k) { /* We must pivot, interchange the rows of the design matrix. */ for (j = 0, pj = pivot*n, kj = k*n; j < n; j++, pj++, kj++) { dtemp = lu[pj]; lu[pj] = lu[kj]; lu[kj] = dtemp; } /* Amend the vector of row maxima. */ dtemp = rowmax[pivot]; rowmax[pivot] = rowmax[k]; rowmax[k] = dtemp; /* Record the interchange for later use. */ itemp = mxl[pivot]; mxl[pivot] = mxl[k]; mxl[k] = itemp; } /* Gaussian elimination. */ for (i = k+1; i < n; i++) { ik = i*n + k; /* Nothing to do if lu[ik] is zero. */ if (lu[ik] != 0.0) { /* Save the scaling factor. */ lu[ik] /= lu[k*n+k]; /* Subtract rows. */ for (j = k+1; j < n; j++) { lu[i*n+j] -= lu[ik]*lu[k*n+j]; } } } } /* mxl[i] records which row of mat corresponds to row i of lu. */ /* lxm[i] records which row of lu corresponds to row i of mat. */ for (i = 0; i < n; i++) { lxm[mxl[i]] = i; } /* Determine the inverse matrix. */ for (i = 0, ij = 0; i < n; i++) { for (j = 0; j < n; j++, ij++) { inv[ij] = 0.0; } } for (k = 0; k < n; k++) { inv[lxm[k]*n+k] = 1.0; /* Forward substitution. */ for (i = lxm[k]+1; i < n; i++) { for (j = lxm[k]; j < i; j++) { inv[i*n+k] -= lu[i*n+j]*inv[j*n+k]; } } /* Backward substitution. */ for (i = n-1; i >= 0; i--) { for (j = i+1; j < n; j++) { inv[i*n+k] -= lu[i*n+j]*inv[j*n+k]; } inv[i*n+k] /= lu[i*n+i]; } } free(mxl); free(lxm); free(rowmax); free(lu); return 0; }