jquant2.c

EVM板JPEG实现,Texas Instruments TMS320C54x EVM JPEG

  /* Note it is important to get the rounding correct! */
  histptr histp;
  int c0,c1,c2;
  int c0min,c0max,c1min,c1max,c2min,c2max;
  long count;
  long total = 0;
  long c0total = 0;
  long c1total = 0;
  long c2total = 0;
  
  c0min = boxp->c0min;  c0max = boxp->c0max;
  c1min = boxp->c1min;  c1max = boxp->c1max;
  c2min = boxp->c2min;  c2max = boxp->c2max;
  
  for (c0 = c0min; c0 <= c0max; c0++)
    for (c1 = c1min; c1 <= c1max; c1++) {
      histp = & histogram[c0][c1][c2min];
      for (c2 = c2min; c2 <= c2max; c2++) {
	if ((count = *histp++) != 0) {
	  total += count;
	  c0total += ((c0 << Y_SHIFT) + ((1<<Y_SHIFT)>>1)) * count;
	  c1total += ((c1 << C_SHIFT) + ((1<<C_SHIFT)>>1)) * count;
	  c2total += ((c2 << C_SHIFT) + ((1<<C_SHIFT)>>1)) * count;
	}
      }
    }
  
  my_colormap[0][icolor] = (JSAMPLE) ((c0total + (total>>1)) / total);
  my_colormap[1][icolor] = (JSAMPLE) ((c1total + (total>>1)) / total);
  my_colormap[2][icolor] = (JSAMPLE) ((c2total + (total>>1)) / total);
}


LOCAL void
remap_colormap (decompress_info_ptr cinfo)
/* Remap the internal colormap to the output colorspace */
{
  /* This requires a little trickery since color_convert expects to
   * deal with 3-D arrays (a 2-D sample array for each component).
   * We must promote the colormaps into one-row 3-D arrays.
   */
  short ci;
  JSAMPARRAY input_hack[3];
  JSAMPARRAY output_hack[10];	/* assume no more than 10 output components */

  for (ci = 0; ci < 3; ci++)
    input_hack[ci] = &(my_colormap[ci]);
  for (ci = 0; ci < cinfo->color_out_comps; ci++)
    output_hack[ci] = &(cinfo->colormap[ci]);

  (*cinfo->methods->color_convert) (cinfo, 1,
				    (long) cinfo->actual_number_of_colors,
				    input_hack, output_hack);
}


LOCAL void
select_colors (decompress_info_ptr cinfo)
/* Master routine for color selection */
{
  int desired = cinfo->desired_number_of_colors;
  int i;

  /* Allocate workspace for box list */
  boxlist = (boxptr) (*cinfo->emethods->alloc_small) (desired * SIZEOF(box));
  /* Initialize one box containing whole space */
  numboxes = 1;
  boxlist[0].c0min = 0;
  boxlist[0].c0max = MAXJSAMPLE >> Y_SHIFT;
  boxlist[0].c1min = 0;
  boxlist[0].c1max = MAXJSAMPLE >> C_SHIFT;
  boxlist[0].c2min = 0;
  boxlist[0].c2max = MAXJSAMPLE >> C_SHIFT;
  /* Shrink it to actually-used volume and set its statistics */
  update_box(& boxlist[0]);
  /* Perform median-cut to produce final box list */
  median_cut(desired);
  /* Compute the representative color for each box, fill my_colormap[] */
  for (i = 0; i < numboxes; i++)
    compute_color(& boxlist[i], i);
  cinfo->actual_number_of_colors = numboxes;
  /* Produce an output colormap in the desired output colorspace */
  remap_colormap(cinfo);
/*  TRACEMS1(cinfo->emethods, 1, "Selected %d colors for quantization",
      numboxes); */
  send_command(ERR5);
  /* Done with the box list */
  (*cinfo->emethods->free_small) ((void *) boxlist);
}


/*
 * These routines are concerned with the time-critical task of mapping input
 * colors to the nearest color in the selected colormap.
 *
 * We re-use the histogram space as an "inverse color map", essentially a
 * cache for the results of nearest-color searches.  All colors within a
 * histogram cell will be mapped to the same colormap entry, namely the one
 * closest to the cell's center.  This may not be quite the closest entry to
 * the actual input color, but it's almost as good.  A zero in the cache
 * indicates we haven't found the nearest color for that cell yet; the array
 * is cleared to zeroes before starting the mapping pass.  When we find the
 * nearest color for a cell, its colormap index plus one is recorded in the
 * cache for future use.  The pass2 scanning routines call fill_inverse_cmap
 * when they need to use an unfilled entry in the cache.
 *
 * Our method of efficiently finding nearest colors is based on the "locally
 * sorted search" idea described by Heckbert and on the incremental distance
 * calculation described by Spencer W. Thomas in chapter III.1 of Graphics
 * Gems II (James Arvo, ed.  Academic Press, 1991).  Thomas points out that
 * the distances from a given colormap entry to each cell of the histogram can
 * be computed quickly using an incremental method: the differences between
 * distances to adjacent cells themselves differ by a constant.  This allows a
 * fairly fast implementation of the "brute force" approach of computing the
 * distance from every colormap entry to every histogram cell.  Unfortunately,
 * it needs a work array to hold the best-distance-so-far for each histogram
 * cell (because the inner loop has to be over cells, not colormap entries).
 * The work array elements have to be INT32s, so the work array would need
 * 256Kb at our recommended precision.  This is not feasible in DOS machines.
 * Another disadvantage of the brute force approach is that it computes
 * distances to every cell of the cubical histogram.  When working with YCbCr
 * input, only about a quarter of the cube represents realizable colors, so
 * many of the cells will never be used and filling them is wasted effort.
 *
 * To get around these problems, we apply Thomas' method to compute the
 * nearest colors for only the cells within a small subbox of the histogram.
 * The work array need be only as big as the subbox, so the memory usage
 * problem is solved.  A subbox is processed only when some cell in it is
 * referenced by the pass2 routines, so we will never bother with cells far
 * outside the realizable color volume.  An additional advantage of this
 * approach is that we can apply Heckbert's locality criterion to quickly
 * eliminate colormap entries that are far away from the subbox; typically
 * three-fourths of the colormap entries are rejected by Heckbert's criterion,
 * and we need not compute their distances to individual cells in the subbox.
 * The speed of this approach is heavily influenced by the subbox size: too
 * small means too much overhead, too big loses because Heckbert's criterion
 * can't eliminate as many colormap entries.  Empirically the best subbox
 * size seems to be about 1/512th of the histogram (1/8th in each direction).
 *
 * Thomas' article also describes a refined method which is asymptotically
 * faster than the brute-force method, but it is also far more complex and
 * cannot efficiently be applied to small subboxes.  It is therefore not
 * useful for programs intended to be portable to DOS machines.  On machines
 * with plenty of memory, filling the whole histogram in one shot with Thomas'
 * refined method might be faster than the present code --- but then again,
 * it might not be any faster, and it's certainly more complicated.
 */


#ifndef BOX_Y_LOG		/* so you can override from Makefile */
#define BOX_Y_LOG  (HIST_Y_BITS-3) /* log2(hist cells in update box, Y axis) */
#endif
#ifndef BOX_C_LOG		/* so you can override from Makefile */
#define BOX_C_LOG  (HIST_C_BITS-3) /* log2(hist cells in update box, C axes) */
#endif

#define BOX_Y_ELEMS  (1<<BOX_Y_LOG) /* # of hist cells in update box */
#define BOX_C_ELEMS  (1<<BOX_C_LOG)

#define BOX_Y_SHIFT  (Y_SHIFT + BOX_Y_LOG)
#define BOX_C_SHIFT  (C_SHIFT + BOX_C_LOG)


/*
 * The next three routines implement inverse colormap filling.  They could
 * all be folded into one big routine, but splitting them up this way saves
 * some stack space (the mindist[] and bestdist[] arrays need not coexist)
 * and may allow some compilers to produce better code by registerizing more
 * inner-loop variables.
 */

LOCAL int
find_nearby_colors (decompress_info_ptr cinfo, int minc0, int minc1, int minc2,
		    JSAMPLE colorlist[])
/* Locate the colormap entries close enough to an update box to be candidates
 * for the nearest entry to some cell(s) in the update box.  The update box
 * is specified by the center coordinates of its first cell.  The number of
 * candidate colormap entries is returned, and their colormap indexes are
 * placed in colorlist[].
 * This routine uses Heckbert's "locally sorted search" criterion to select
 * the colors that need further consideration.
 */
{
  int numcolors = cinfo->actual_number_of_colors;
  int maxc0, maxc1, maxc2;
  int centerc0, centerc1, centerc2;
  int i, x, ncolors;
  INT32 minmaxdist, min_dist, max_dist, tdist;
  INT32 mindist[MAXNUMCOLORS];	/* min distance to colormap entry i */

  /* Compute true coordinates of update box's upper corner and center.
   * Actually we compute the coordinates of the center of the upper-corner
   * histogram cell, which are the upper bounds of the volume we care about.
   * Note that since ">>" rounds down, the "center" values may be closer to
   * min than to max; hence comparisons to them must be "<=", not "<".
   */
  maxc0 = minc0 + ((1 << BOX_Y_SHIFT) - (1 << Y_SHIFT));
  centerc0 = (minc0 + maxc0) >> 1;
  maxc1 = minc1 + ((1 << BOX_C_SHIFT) - (1 << C_SHIFT));
  centerc1 = (minc1 + maxc1) >> 1;
  maxc2 = minc2 + ((1 << BOX_C_SHIFT) - (1 << C_SHIFT));
  centerc2 = (minc2 + maxc2) >> 1;

  /* For each color in colormap, find:
   *  1. its minimum squared-distance to any point in the update box
   *     (zero if color is within update box);
   *  2. its maximum squared-distance to any point in the update box.
   * Both of these can be found by considering only the corners of the box.
   * We save the minimum distance for each color in mindist[];
   * only the smallest maximum distance is of interest.
   * Note we have to scale Y to get correct distance in scaled space.
   */
  minmaxdist = 0x7FFFFFFFL;

  for (i = 0; i < numcolors; i++) {
    /* We compute the squared-c0-distance term, then add in the other two. */
    x = GETJSAMPLE(my_colormap[0][i]);
    if (x < minc0) {
      tdist = (x - minc0) * Y_SCALE;
      min_dist = tdist*tdist;
      tdist = (x - maxc0) * Y_SCALE;
      max_dist = tdist*tdist;
    } else if (x > maxc0) {
      tdist = (x - maxc0) * Y_SCALE;
      min_dist = tdist*tdist;
      tdist = (x - minc0) * Y_SCALE;
      max_dist = tdist*tdist;
    } else {
      /* within cell range so no contribution to min_dist */
      min_dist = 0;
      if (x <= centerc0) {
	tdist = (x - maxc0) * Y_SCALE;
	max_dist = tdist*tdist;
      } else {
	tdist = (x - minc0) * Y_SCALE;
	max_dist = tdist*tdist;
      }
    }

    x = GETJSAMPLE(my_colormap[1][i]);
    if (x < minc1) {
      tdist = x - minc1;
      min_dist += tdist*tdist;
      tdist = x - maxc1;
      max_dist += tdist*tdist;
    } else if (x > maxc1) {
      tdist = x - maxc1;
      min_dist += tdist*tdist;
      tdist = x - minc1;
      max_dist += tdist*tdist;
    } else {
      /* within cell range so no contribution to min_dist */
      if (x <= centerc1) {
	tdist = x - maxc1;
	max_dist += tdist*tdist;
      } else {
	tdist = x - minc1;
	max_dist += tdist*tdist;
      }
    }

    x = GETJSAMPLE(my_colormap[2][i]);
    if (x < minc2) {
      tdist = x - minc2;
      min_dist += tdist*tdist;
      tdist = x - maxc2;
      max_dist += tdist*tdist;
    } else if (x > maxc2) {
      tdist = x - maxc2;
      min_dist += tdist*tdist;
      tdist = x - minc2;
      max_dist += tdist*tdist;
    } else {
      /* within cell range so no contribution to min_dist */
      if (x <= centerc2) {
	tdist = x - maxc2;
	max_dist += tdist*tdist;
      } else {
	tdist = x - minc2;
	max_dist += tdist*tdist;
      }
    }

    mindist[i] = min_dist;	/* save away the results */
    if (max_dist < minmaxdist)
      minmaxdist = max_dist;
  }

  /* Now we know that no cell in the update box is more than minmaxdist
   * away from some colormap entry.  Therefore, only colors that are
   * within minmaxdist of some part of the box need be considered.
   */
  ncolors = 0;
  for (i = 0; i < numcolors; i++) {
    if (mindist[i] <= minmaxdist)
      colorlist[ncolors++] = (JSAMPLE) i;
  }
  return ncolors;
}


LOCAL void
find_best_colors (decompress_info_ptr cinfo, int minc0, int minc1, int minc2,
		  int numcolors, JSAMPLE colorlist[], JSAMPLE bestcolor[])
/* Find the closest colormap entry for each cell in the update box,
 * given the list of candidate colors prepared by find_nearby_colors.
 * Return the indexes of the closest entries in the bestcolor[] array.
 * This routine uses Thomas' incremental distance calculation method to
 * find the distance from a colormap entry to successive cells in the box.
 */
{
  int ic0, ic1, ic2;
  int i, icolor;
  register INT32 * bptr;	/* pointer into bestdist[] array */
  JSAMPLE * cptr;		/* pointer into bestcolor[] array */
  INT32 dist0, dist1;		/* initial distance values */
  register INT32 dist2;		/* current distance in inner loop */
  INT32 xx0, xx1;		/* distance increments */
  register INT32 xx2;
  INT32 inc0, inc1, inc2;	/* initial values for increments */
  /* This array holds the distance to the nearest-so-far color for each cell */
  INT32 bestdist[BOX_Y_ELEMS * BOX_C_ELEMS * BOX_C_ELEMS];

  /* Initialize best-distance for each cell of the update box */
  bptr = bestdist;
  for (i = BOX_Y_ELEMS*BOX_C_ELEMS*BOX_C_ELEMS-1; i >= 0; i--)
    *bptr++ = 0x7FFFFFFFL;
  
  /* For each color selected by find_nearby_colors,
   * compute its distance to the center of each cell in the box.
   * If that's less than best-so-far, update best distance and color number.
   * Note we have to scale Y to get correct distance in scaled space.
   */
  
  /* Nominal steps between cell centers ("x" in Thomas article) */
#define STEP_Y  ((1 << Y_SHIFT) * Y_SCALE)
#define STEP_C  (1 << C_SHIFT)
  
  for (i = 0; i < numcolors; i++) {
    icolor = GETJSAMPLE(colorlist[i]);
    /* Compute (square of) distance from minc0/c1/c2 to this color */
    inc0 = (minc0 - (int) GETJSAMPLE(my_colormap[0][icolor])) * Y_SCALE;
    dist0 = inc0*inc0;
    inc1 = minc1 - (int) GETJSAMPLE(my_colormap[1][icolor]);
    dist0 += inc1*inc1;
    inc2 = minc2 - (int) GETJSAMPLE(my_colormap[2][icolor]);
    dist0 += inc2*inc2;
    /* Form the initial difference increments */
    inc0 = inc0 * (2 * STEP_Y) + STEP_Y * STEP_Y;
    inc1 = inc1 * (2 * STEP_C) + STEP_C * STEP_C;
    inc2 = inc2 * (2 * STEP_C) + STEP_C * STEP_C;
    /* Now loop over all cells in box, updating distance per Thomas method */
    bptr = bestdist;
    cptr = bestcolor;
    xx0 = inc0;
    for (ic0 = BOX_Y_ELEMS-1; ic0 >= 0; ic0--) {
      dist1 = dist0;
      xx1 = inc1;
      for (ic1 = BOX_C_ELEMS-1; ic1 >= 0; ic1--) {
	dist2 = dist1;
	xx2 = inc2;
	for (ic2 = BOX_C_ELEMS-1; ic2 >= 0; ic2--) {
	  if (dist2 < *bptr) {
	    *bptr = dist2;
	    *cptr = (JSAMPLE) icolor;
	  }
	  dist2 += xx2;
	  xx2 += 2 * STEP_C * STEP_C;
	  bptr++;
	  cptr++;
	}
	dist1 += xx1;
	xx1 += 2 * STEP_C * STEP_C;
      }
      dist0 += xx0;
      xx0 += 2 * STEP_Y * STEP_Y;
    }
  }
}


LOCAL void
fill_inverse_cmap (decompress_info_ptr cinfo, int c0, int c1, int c2)
/* Fill the inverse-colormap entries in the update box that contains */
/* histogram cell c0/c1/c2.  (Only that one cell MUST be filled, but */
/* we can fill as many others as we wish.) */
{
  int minc0, minc1, minc2;	/* lower left corner of update box */
  int ic0, ic1, ic2;
  register JSAMPLE * cptr;	/* pointer into bestcolor[] array */
  register histptr cachep;	/* pointer into main cache array */
  /* This array lists the candidate colormap indexes. */
  JSAMPLE colorlist[MAXNUMCOLORS];
  int numcolors;		/* number of candidate colors */
  /* This array holds the actually closest colormap index for each cell. */