im_matmul(3) matrix operations on DOUBLEMASKs

Other Alias

im_matinv, im_mattrn


#include <vips/vips.h>

DOUBLEMASK *im_matinv( const DOUBLEMASK *in, const char *name );

int im_matinv_inplace( DOUBLEMASK *mat );

DOUBLEMASK *im_matmul( in1, in2, name )
DOUBLEMASK *in1, *in2;
char *name;

DOUBLEMASK *im_matcat( in1, in2, name )
DOUBLEMASK *in1, *in2;
char *name;

DOUBLEMASK *im_mattrn( in, name )
char *name;


These functions treat DOUBLEMASKs as matricies, performing some of the basics of matrix algebra on them.

There should be more matrix functions: those implemeneted are just sufficient for the Gallery's calibration routines. im_matadd, im_matidentity should really be added.

None of these functions damage their arguments, except im_matinv_inplace(3).

im_matinv(3) inverts DOUBLEMASK in, returning a new DOUBLEMASK, called name, which contains the inverse of in. If no inverse exists, NULL is returned and im_error(3) is called with a diagnostic message.

im_matinv_inplace(3) is as im_matinv(3) except that it overwrites its input.

im_matmul() multiples the matrices held in in1 and in2, returning their product in a matrix called name.

im_matcat() returns a new matrix formed by appending matrix in2 to the end of matrix in1. The two matricies must be the same width. It is useful for combining several im_measure()s into a single matrix.

im_mattrn() transposes matrix in, returning the transpose in new matrix called name.


DO NOT use matrix inversion to solve systems of linear equations (SLEs). The routines im_lu_decomp(3) and im_lu_solve(3) are more efficient, even for a single SLE.


The functions returns a new DOUBLEMASK on sucess, and NULL on failure.

im_matinv_inplace(3) returns zero on success, and -1 on failure.


im_create_dmask(3), im_measure(3), etc. im_lu_decomp(3), im_lu_solve(3)


National Gallery, 1992. Tom Vajzovic, 2006


J. Cupitt
Tom Vajzovic