new_ms_eigsearch.c

FERET人脸库的处理代码。内函预处理

/*----------------------------------------------------------------------
PROGRAM: new_ms_eigsearch.c
DATE:    2/24/95
AUTHOR:  Baback Moghaddam, [e-mail]
------------------------------------------------------------------------


  Multiple-Scale Local feature search by eigentemplates

  This routine looks for 2 (3) BF files in the datapath (defined below):

  1. features.bf  which is an N-by-2 matrix defining the size (row,col)
     of each of the N templates.

  2. template[n].bf  where n=1...N is the eigenvector ROW matrix where
     the 1st row is the mean feature, and the remaining rows
     are the principal eigenvectors 1:M. 

     NOTE: The eigenvectors are stored in column-order (as in MATLAB)

  3. variances.bf which is an N-by-M matrix, where each row represents
     the rank-ordered eigenvalues associated with that feature


     NOTE: The set of scales over which the search is to be performed
           is provided in an ASCII file of the format:

	   s1
	   s2
	   .
	   .
	   .
	   sN

----------------------------------------------------------------------

   NOTE: In this version if unitmap'ing is selected it will use
         the Mahalanobis mean/stddev in addition to the standard
	 distance computation.


	 Also fixed the unitmap musigma cost 

	 Includes patch dump additions from rena

------------------------------------------------------------------------

     Configure this code using the following defines:
     
     DO_VAR:     will set each pixel whose surrond's variance
                 is < VAR_THRESHOLD to VAR_MAXVALUE
	       
---------------------------------------------------------------------- */

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <math.h>
#include <float.h>
#include "util.h"
#include "io.h"
#include "matrix.h"
#include "affine.h"

struct Points {
  int n;
  int *x;
  int *y;
  float *f;
};

struct Point {
  int x;
  int y;
  float f;
};


/* ----------- CONFIGURES ---------------- */

#define DO_VAR                  0     /*  variance check     */
#define VAR_THRESHOLD           100   /*  threshold for low-var patches */
#define VAR_MAXVALUE            1e7   /*  replacement in distance map   */

#define MAX_NUM_TEMPLATES       5
#define MAX_NUM_SCALES          32
#define MAX_NUM_EIGENVECTORS    128   /* for static storage in dffs() */
#define MAX_VECTOR_LENGTH       128    /* for static storage in mahalanobis_distance() */
#define MAX_CHARS               512

/* ----------- Command-Line Parsing Stuff ------- */

extern int optind;
extern char *optarg;
char *progname;          /* used to store the name of the program  */
char comline[MAX_CHARS]; /* used to store the entire command line  */

#define OPTIONS "i:l:s:d:o:a:b:k:gumnvcefp"      /* RENA added p option */

char *usage = "\t-i indir -l list -s scalefile [-d datadir]\n\
\t\t\t[-o outfile] [-a first_ev] [-b last_ev] [-k feature]\n\
\t\t\t[-g] [-u] [-m] [-n] [-v] [-e] [-c] [-f] [-p]\n";       /* RENA added [-p] */

char *help = "\
Multiscale Eigentemplate Search\n\n\
-i indir   \t input directory\n\
-l listfile\t ASCII file listing indir files to process (one per line)\n\
-s scalefile\t ASCII file of scales\n\
-d datadir \t data dir holding eigentemplate BF files (default = ./)\n\
-o outfile \t output file (default = ./ms_eigsearch.out)\n\
-a first_ev\t first eigenvector (default = 1)\n\
-b last_ev \t last eigenvector  (default = 5)\n\
-k feature \t search for the k-th feature only\n\
-g         \t pre-process with graymap\n\
-u         \t pre-process with unitmap\n\
-m         \t use Mahalanobis distance instead of DFFS\n\
-n         \t use optimal weight rho^* (default = lambda_{last_ev+1})\n\
-v         \t use normalized correlation (distance = 1 - nc)\n\
-c         \t use spatial priors\n\
-e         \t output detection maps (in indir/Maps)\n\
-f         \t output ASCII minima files to outdir\n\
-p         \t output image patches to outdir\n";          /* RENA added -p help description */


/* --------- Function Prototypes ---------- */

float dffs(float *patch, int N, float **eigvectors, int first, int last);
float mahalanobis(float *patch, int N, float **eigvectors, float *eigvalues,
		  float lambda_star, int first, int last);
float graymap(float *patch, int N);
float unitmap(float *patch, int N);
void statistics(float *p, int N, float *mean, float *sigma);
float umap_musigma_distance(float mean, float sigma, float **umap_musigma);
float mahalanobis_distance(float *x, int N, float *Mu, float **inv_Sigma);
void minima(float **A, int imin, int imax, int jmin, int jmax,
            struct Points *Ps);
float norm_corr(float *patch, int N,
		float *template, float t_mean, float t_sigma);



/* -------- Globals ---------------- */

float **template[MAX_NUM_TEMPLATES];
float **errormap[MAX_NUM_TEMPLATES];
float **variances;
float **umap_musigma[MAX_NUM_TEMPLATES];
float **spatial_musigma[MAX_NUM_TEMPLATES];


/* ------- Command Line Defaults ------------- */

int do_dmdumps = 0; /* default is NOT to do errormap dumps  */
int do_graymap = 0; /* default is NOT to graymap each patch */
int do_unitmap = 0; /* default is NOT to unitmap each patch */
int do_mahalanobis = 0; /* default is NOT to do this computation */
int do_optimal_rho = 0; /* default is the old Mahalanobis computation */
int do_spatial = 0; /* default is NOT to add [x,y] Mahalanobis dist */
int do_minima  = 0;  /* default is not to output minima files */
int do_correlation = 0; /* default is not to do normalized correlation */
/* RENA begin */
int do_patch_dump = 0; /* default is not to do patch dumps */
/* RENA end */
int eig_first = 1;  /* search using first 5 eigenvectors by default */
int eig_last  = 5;



/*----------------------------------------------------------------------*/
/* ---------------------------- MAIN ---------------------------------- */

main(int argc, char *argv[])
{

  register int i,j,k,l,ii,jj;
  int f,s,c,feature=1,nframe,nfeatures,sets, bytes_pixel;
  int nrow, ncol, max_N, xc, yc;
  char command[MAX_CHARS],indir[MAX_CHARS],infile[MAX_CHARS];
  char listfile[MAX_CHARS], scalefile[MAX_CHARS], line[MAX_CHARS];
  char datapath[MAX_CHARS],outdir[MAX_CHARS],filename[MAX_CHARS];
  char outfile[MAX_CHARS];

  float **image, **image_orig;
  float fval1, fval2, fval, mean, sigma;
  float maxerror;
  FILE *fp, *fp1, *fp2;         /* for output values dump */

  int rowl[MAX_NUM_TEMPLATES],rowr[MAX_NUM_TEMPLATES];
  int coll[MAX_NUM_TEMPLATES],colr[MAX_NUM_TEMPLATES];
  int imin[MAX_NUM_TEMPLATES],imax[MAX_NUM_TEMPLATES];
  int jmin[MAX_NUM_TEMPLATES],jmax[MAX_NUM_TEMPLATES];
  int N_dim[MAX_NUM_TEMPLATES]; /* the dimensionality of each template */
  int M_dim[MAX_NUM_TEMPLATES]; /* the # of eigenvectors for each template */

  int numscales;
  float scales[MAX_NUM_SCALES];
  int rowm[MAX_NUM_SCALES][MAX_NUM_TEMPLATES];
  int colm[MAX_NUM_SCALES][MAX_NUM_TEMPLATES];
  float error[MAX_NUM_SCALES][MAX_NUM_TEMPLATES];
  float **Warp;
  float *my_vector;
  int xdim, ydim;
  int   best_scale[MAX_NUM_TEMPLATES];
  float best_error[MAX_NUM_TEMPLATES];
  int   best_rowm[MAX_NUM_TEMPLATES];
  int   best_colm[MAX_NUM_TEMPLATES];

  
  int **templatesize;
  float *patch;
  /* RENA begin */
  unsigned char **image_patch[MAX_NUM_TEMPLATES];
  /* RENA end */

  struct Points myPoints_array[MAX_NUM_TEMPLATES];
  int N_minima;

  float lambda_star[MAX_NUM_TEMPLATES];     /* optimal variance for DFFS */
  float template_mean[MAX_NUM_TEMPLATES];   /* statistics for norm_corr */
  float template_sigma[MAX_NUM_TEMPLATES];  /* statistics for norm_corr */

  /* required input flags */
  
  int errflag   = 0;
  int inflag    = 0;
  int scaleflag = 0;
  int listflag  = 0;
  

  /* command line defaults */

  int singlefeature = 0;  
  int firstframe = 0;
  int lastframe  = 0;
  strcpy(datapath,".");               /* default datapath directory to cwd   */


  /* setup program name and command line strings */
  
  progname = argv[0];
  for (i=0; i<argc; i++)
    strcat(comline, argv[i]),strcat(comline, " ");


  /* set up default outfile name */

  sprintf(outfile,"%s.out",progname); /* default output file is progname.out */  

  /* ----------------------  Command Line Parse ------------------------ */
  
  while ((c = getopt(argc, argv, OPTIONS)) != EOF)
    switch (c) {
      
    case 'i':
      strcpy(indir, optarg);
      inflag = 1;
      break;
      
    case 'l':
      strcpy(listfile, optarg);
      listflag = 1;
      break;

    case 'd':
      strcpy(datapath, optarg);
      break;

    case 's':
      strcpy(scalefile, optarg);
      scaleflag = 1;
      break;

    case 'k':
      feature = atoi(optarg);
      singlefeature = 1;
      break;

    case 'o':
      strcpy(outfile, optarg);
      break;

    case 'a':
      eig_first  = atoi(optarg);
      break;

    case 'b':
      eig_last = atoi(optarg);
      break;

    case 'g':
      do_graymap = 1;
      break;

    case 'u':
      do_unitmap = 1;
      break;

    case 'e':
      do_dmdumps = 1;
      break;

    case 'f':
      do_minima = 1;
      break;

    case 'm':
      do_mahalanobis = 1;
      break;
 
    case 'n':
      do_optimal_rho = 1;
      break;

    case 'v':
      do_correlation = 1;
      break;
 
    case 'c':
      do_spatial = 1;
      break;

    /* RENA begin */
    case 'p':
      do_patch_dump = 1;
      break;
    /* RENA end */

    case '?':
      errflag = 1;
      break;
      
    }
  
  
  /* command line error check */
  
  if (errflag || !inflag || !scaleflag || !listflag) {
    fprintf(stderr,"\nUSAGE: %s %s\n%s\n", progname, usage, help);
    exit(1);
  }

  sprintf(outdir,"%s/Maps",indir);   /* default output dir is indir/Maps    */



  /* -------- Load eigentemplate data --------------- */

  sprintf(filename,"%s/features.bf",datapath);
 
  { 
    float **m;
    max_N = 0;
    m = read_BIN(filename, &nfeatures, &ncol);
    templatesize = imatrix(1, nfeatures, 1, 2);
    
    for (i=1; i<=nfeatures; i++) {
      for (j=1; j<=ncol; j++)
	templatesize[i][j] = (int) m[i][j];
      N_dim[i] = templatesize[i][1]*templatesize[i][2];
      if (N_dim[i]>max_N) 
	max_N = N_dim[i];
      free_matrix(m, 1, nfeatures, 1, ncol);
      /* RENA begin */
      if (do_patch_dump) {
	image_patch[i] = cmatrix(1, templatesize[i][1], 1, templatesize[i][2]);
	/* RENA end */
      }
    }
  }
  
  for (i=1; i<=nfeatures; i++) {
    sprintf(filename,"%s/template%d.bf",datapath,i);
    template[i] = read_BIN(filename, &nrow, &ncol);
    if (do_correlation) {
      statistics(template[i][1], ncol, &template_mean[i], &template_sigma[i]);
      fprintf(stdout,"Template-%d \t mean = %f \t sigma = %f\n\n",
	      i, template_mean[i], template_sigma[i]);
    }
    M_dim[i] = nrow;
    if (ncol!=N_dim[i])
      myerror("Template sizes in BF don't match those is definition file");
  }

  if (do_mahalanobis) {
    sprintf(filename,"%s/variances.bf",datapath);
    variances = read_BIN(filename, &nrow, &ncol);
    if (nrow!=nfeatures) {
      fprintf(stderr,
	      "ERROR: variances.bf has %d rows! (must have %d)\n\n",
	      nrow, nfeatures);
      exit(1);
    }
    if (eig_last>ncol) {
      fprintf(stderr,
	      "ERROR: variances.bf has %d cols! (must have atleast %d)\n\n",
	      ncol, eig_last);
      exit(1);
    }

    /* now estimate lambda_star */
    
    for (i=1; i<=nfeatures; i++) {
      if (do_optimal_rho>0) {
	/* average eigenvalues in orthogonal subspace */
	lambda_star[i] = 0.0;
	for (j=eig_last+1; j<=ncol-1; j++)  /* sum to next to last in case <=0 */
	  lambda_star[i] += variances[i][j];
	lambda_star[i] += ((N_dim[i]-ncol+1) * variances[i][ncol-1]);
	lambda_star[i] /= (N_dim[i] - eig_last);
      }
      else {
	lambda_star[i] = variances[i][eig_last+1];
      }
    }
  }
  
  if (do_unitmap) {
    float **C, **invC;
    int m, n;
    C = matrix(1,2,1,2);
    invC = matrix(1,2,1,2);
    for (k=1; k<=nfeatures; k++) {
      sprintf(filename,"%s/umap_musigma%d.bf",datapath,k);
      umap_musigma[k] = read_BIN(filename, &m, &n );
     
      /* invert the covariance matrix portion of umap_musigma[k] */
      for (i=1; i<=2; i++)
	for (j=1; j<=2; j++)
	  C[i][j] = umap_musigma[k][i+1][j];
      matrix_inverse(C, 2, invC);
      for (i=1; i<=2; i++)
	for (j=1; j<=2; j++)
	  umap_musigma[k][i+1][j] = invC[i][j];

      /* ----------- DEBUG -----------
      for (i=1; i<=3; i++) {
	for (j=1; j<=2; j++)
	  printf("%f ",umap_musigma[k][i][j]);
	printf("\n");
      }
      printf("\n\n");
      C[1][1] = 10;
      C[1][2] = 10;
      printf("dist = %f \n", 
	     mahalanobis_distance(C[1], 2, umap_musigma[k][1], &umap_musigma[k][1]));
      printf("dist = %f \n",
		    umap_musigma_distance(10, 10, umap_musigma[k]));
      exit(1);
      --------------- DEBUG ------------- */
    }  
    free_matrix(C,1,2,1,2);
    free_matrix(invC,1,2,1,2);
  }

   
  if (do_spatial) {
    float **C, **invC;
    int m, n;
    C = matrix(1,2,1,2);
    invC = matrix(1,2,1,2);
    for (k=1; k<=nfeatures; k++) {
      sprintf(filename,"%s/spatial_musigma%d.bf",datapath,k);
      spatial_musigma[k] = read_BIN(filename, &m, &n );
     
      /* invert the covariance matrix portion of umap_musigma[k] */
      for (i=1; i<=2; i++)
	for (j=1; j<=2; j++)
	  C[i][j] = spatial_musigma[k][i+1][j];
      matrix_inverse(C, 2, invC);
      for (i=1; i<=2; i++)
	for (j=1; j<=2; j++)
	  spatial_musigma[k][i+1][j] = invC[i][j];

      /* ----------- DEBUG -----------
      for (i=1; i<=3; i++) {
	for (j=1; j<=2; j++)
	  printf("%f ",spatial_musigma[k][i][j]);
	printf("\n");
      }
      printf("\n\n");
      C[1][1] = 10;
      C[1][2] = 10;
      printf("dist = %f \n", 
	     mahalanobis_distance(C[1], 2, spatial_musigma[k][1], &spatial_musigma[k][1]));
      printf("dist = %f \n",
		    spatial_musigma_distance(10, 10, spatial_musigma[k]));
      exit(1);
      --------------- DEBUG ------------- */
    }  
    free_matrix(C,1,2,1,2);
    free_matrix(invC,1,2,1,2);
  }

  /* ---- determine left/right boundaries for errormaps ----- */
  for (i=1; i<=nfeatures; i++) {
    rowl[i] = (int) templatesize[i][1]/2.0;
    rowr[i] = templatesize[i][1] - rowl[i] - 1;
    coll[i] = (int) templatesize[i][2]/2.0;
    colr[i] = templatesize[i][2] - coll[i] - 1;
  }

  patch = vector(1, max_N);
  my_vector = vector(1, MAX_VECTOR_LENGTH);


  /* ----  read indir descriptor -------- */
  
  read_descriptor(indir, &nframe, &sets, &bytes_pixel, &ncol, &nrow);
  if (sets>1) 
    myerror("Input files must be single-set DAT files!");
  image = matrix(1, nrow, 1, ncol);
  image_orig = matrix(1, nrow, 1, ncol);


  /* --- allocate errormap tensor and set window boundaries --- */

  for (k=feature; k<=(singlefeature>0 ? feature:nfeatures); k++) {
    errormap[k] = matrix(1, nrow, 1, ncol);
    imin[k] = rowl[k] + 1;
    imax[k] = nrow - rowr[k];
    jmin[k] = coll[k] + 1;
    jmax[k] = ncol - colr[k];
  }
  
  if (do_minima) {
    int i1 = imin[feature], i2 = imax[feature];
    int j1 = jmin[feature], j2 = jmax[feature];
    int numpixels = (i2-i1)*(j2-j1), n;
    
    for (k=feature; k<=(singlefeature>0 ? feature:nfeatures); k++) {
      n = (imax[k]-imin[k])*(jmax[k]-jmin[k]);
      if (n>numpixels) {
	numpixels = n;
	i1 = imin[k];
	i2 = imax[k];
	j1 = jmin[k];
	j2 = jmax[k];
      }
    }
    N_minima = (i2-i1)*(j2-j1) / 9;
    for (k=feature; k<=(singlefeature>0 ? feature:nfeatures); k++) {
      myPoints_array[k].n = N_minima;
      myPoints_array[k].x = ivector(1, N_minima);
      myPoints_array[k].y = ivector(1, N_minima);
      myPoints_array[k].f =  vector(1, N_minima);
    }
  }

  /* -------- Load scalefile -------------- */

  sprintf(filename,"%s",scalefile);
  if ((fp = fopen(filename, "r")) == NULL) {
    fprintf(stderr,"ERROR Could not open scale file %s \n\n", filename);
    exit(1);
  }
  i = 1;
  while (fgets(line, MAX_CHARS, fp)) {
    if (strncmp(line, "#", 1) != 0 && strlen(line)>1) {
      sscanf(line, "%f", scales+i);
      i++;
    }
  }
  numscales = i-1;
  fclose(fp);