pointop.c

基于小波域隐马尔可夫模型的图像降噪

/* 
RES = pointOp(IM, LUT, ORIGIN, INCREMENT, WARNINGS)
  >>> See pointOp.m for documentation <<<
  EPS, ported from OBVIUS, 7/96.
*/

/* Matlab V4 types should be changed as follows:
  Matrix -> mxArray
  REAL -> mxREAL
  mxCreateFull ->  mxCreateDoubleMatrix
  */

#define V4_COMPAT
#include <matrix.h>  /* Matlab matrices */
#include <mex.h>

#include <stddef.h>  /* NULL */

#define notDblMtx(it) (!mxIsNumeric(it) || !mxIsDouble(it) || mxIsSparse(it) || mxIsComplex(it))

void internal_pointop();

void mexFunction(int nlhs,	     /* Num return vals on lhs */
		 Matrix *plhs[],    /* Matrices on lhs      */
		 int nrhs,	     /* Num args on rhs    */
		 Matrix *prhs[]     /* Matrices on rhs */
		 )
  {
  double *image, *lut, *res;
  double origin, increment;
  int x_dim, y_dim, lx_dim, ly_dim;
  int warnings = 1;
  Matrix *arg;
  double *mxMat;

  if (nrhs < 4 ) mexErrMsgTxt("requres  at least 4 args.");

  /* ARG 1: IMAGE  */
  arg = prhs[0];
  if notDblMtx(arg) mexErrMsgTxt("IMAGE arg must be a real non-sparse matrix.");
  image = mxGetPr(arg);
  x_dim = (int) mxGetM(arg); /* X is inner index! */
  y_dim = (int) mxGetN(arg);

  /* ARG 2: Lookup table */
  arg = prhs[1];
  if notDblMtx(arg) mexErrMsgTxt("LUT arg must be a real non-sparse matrix.");
  lut = mxGetPr(arg);
  lx_dim = (int) mxGetM(arg); /* X is inner index! */
  ly_dim = (int) mxGetN(arg);
  if ( (lx_dim != 1) && (ly_dim != 1) )
    mexErrMsgTxt("Lookup table must be a row or column vector.");

  /* ARG 3: ORIGIN */
  arg = prhs[2];
  if notDblMtx(arg) mexErrMsgTxt("ORIGIN arg must be a real scalar.");
  if (mxGetM(arg) * mxGetN(arg) != 1)
     mexErrMsgTxt("ORIGIN arg must be a real scalar.");
  mxMat = mxGetPr(arg);
  origin = *mxMat;

  /* ARG 4: INCREMENT */
  arg = prhs[3];
  if notDblMtx(arg) mexErrMsgTxt("INCREMENT arg must be a real scalar.");
  if (mxGetM(arg) * mxGetN(arg) != 1)
     mexErrMsgTxt("INCREMENT arg must be a real scalar.");
  mxMat = mxGetPr(arg);
  increment = *mxMat;

  /* ARG 5: WARNINGS */
  if (nrhs>4)
    {
    arg = prhs[4];
    if notDblMtx(arg) mexErrMsgTxt("WARINGS arg must be a real scalar.");
    if (mxGetM(arg) * mxGetN(arg) != 1)
      mexErrMsgTxt("WARNINGS arg must be a real scalar.");
    mxMat = mxGetPr(arg);
    warnings = (int) *mxMat;
    }

  plhs[0] = (Matrix *)  mxCreateFull(x_dim,y_dim,REAL);
  if (plhs[0] == NULL) mexErrMsgTxt("Cannot allocate result matrix");
  res = mxGetPr(plhs[0]);
      
  internal_pointop(image, res, x_dim*y_dim, lut, lx_dim*ly_dim, 
		   origin, increment, warnings);
  return;
  }      


/* Use linear interpolation on a lookup table.
   Taken from OBVIUS.  EPS, Spring, 1987.
 */
void internal_pointop (im, res, size, lut, lutsize, origin, increment, warnings)
  register double *im, *res, *lut;
  register double origin, increment; 
  register int size, lutsize, warnings;
  {
  register int i, index;
  register double pos;
  register int l_unwarned = warnings;
  register int r_unwarned = warnings;

  lutsize = lutsize - 2;	/* Maximum index value */
  if (increment > 0)
    for (i=0; i<size; i++)
	{
	pos = (im[i] - origin) / increment;
	index = (int) pos;   /* Floor */
	if (index < 0)
	    {
	    index = 0;
	    if (l_unwarned)
		{
		mexPrintf("Warning: Extrapolating to left of lookup table...\n");
		l_unwarned = 0;
		}
	    }
	else if (index > lutsize)
	    {
	    index = lutsize;
	    if (r_unwarned)
		{
		mexPrintf("Warning: Extrapolating to right of lookup table...\n");
		r_unwarned = 0;
		}
	    }
	res[i] = lut[index] + (lut[index+1] - lut[index]) * (pos - index);
	}
  else
    for (i=0; i<size; i++) res[i] = *lut;
  }