source: trunk/src/3rdparty/libjpeg/jquant2.c@ 561

/*
 * jquant2.c
 *
 * Copyright (C) 1991-1996, Thomas G. Lane.
 * This file is part of the Independent JPEG Group's software.
 * For conditions of distribution and use, see the accompanying README file.
 *
 * This file contains 2-pass color quantization (color mapping) routines.
 * These routines provide selection of a custom color map for an image,
 * followed by mapping of the image to that color map, with optional
 * Floyd-Steinberg dithering.
 * It is also possible to use just the second pass to map to an arbitrary
 * externally-given color map.
 *
 * Note: ordered dithering is not supported, since there isn't any fast
 * way to compute intercolor distances; it's unclear that ordered dither's
 * fundamental assumptions even hold with an irregularly spaced color map.
 */

#define JPEG_INTERNALS
#include "jinclude.h"
#include "jpeglib.h"

#ifdef QUANT_2PASS_SUPPORTED


/*
 * This module implements the well-known Heckbert paradigm for color
 * quantization.  Most of the ideas used here can be traced back to
 * Heckbert's seminal paper
 *   Heckbert, Paul.  "Color Image Quantization for Frame Buffer Display",
 *   Proc. SIGGRAPH '82, Computer Graphics v.16 #3 (July 1982), pp 297-304.
 *
 * In the first pass over the image, we accumulate a histogram showing the
 * usage count of each possible color.  To keep the histogram to a reasonable
 * size, we reduce the precision of the input; typical practice is to retain
 * 5 or 6 bits per color, so that 8 or 4 different input values are counted
 * in the same histogram cell.
 *
 * Next, the color-selection step begins with a box representing the whole
 * color space, and repeatedly splits the "largest" remaining box until we
 * have as many boxes as desired colors.  Then the mean color in each
 * remaining box becomes one of the possible output colors.
 * 
 * The second pass over the image maps each input pixel to the closest output
 * color (optionally after applying a Floyd-Steinberg dithering correction).
 * This mapping is logically trivial, but making it go fast enough requires
 * considerable care.
 *
 * Heckbert-style quantizers vary a good deal in their policies for choosing
 * the "largest" box and deciding where to cut it.  The particular policies
 * used here have proved out well in experimental comparisons, but better ones
 * may yet be found.
 *
 * In earlier versions of the IJG code, this module quantized in YCbCr color
 * space, processing the raw upsampled data without a color conversion step.
 * This allowed the color conversion math to be done only once per colormap
 * entry, not once per pixel.  However, that optimization precluded other
 * useful optimizations (such as merging color conversion with upsampling)
 * and it also interfered with desired capabilities such as quantizing to an
 * externally-supplied colormap.  We have therefore abandoned that approach.
 * The present code works in the post-conversion color space, typically RGB.
 *
 * To improve the visual quality of the results, we actually work in scaled
 * RGB space, giving G distances more weight than R, and R in turn more than
 * B.  To do everything in integer math, we must use integer scale factors.
 * The 2/3/1 scale factors used here correspond loosely to the relative
 * weights of the colors in the NTSC grayscale equation.
 * If you want to use this code to quantize a non-RGB color space, you'll
 * probably need to change these scale factors.
 */

#define R_SCALE 2               /* scale R distances by this much */
#define G_SCALE 3               /* scale G distances by this much */
#define B_SCALE 1               /* and B by this much */

/* Relabel R/G/B as components 0/1/2, respecting the RGB ordering defined
 * in jmorecfg.h.  As the code stands, it will do the right thing for R,G,B
 * and B,G,R orders.  If you define some other weird order in jmorecfg.h,
 * you'll get compile errors until you extend this logic.  In that case
 * you'll probably want to tweak the histogram sizes too.
 */

#if RGB_RED == 0
#define C0_SCALE R_SCALE
#endif
#if RGB_BLUE == 0
#define C0_SCALE B_SCALE
#endif
#if RGB_GREEN == 1
#define C1_SCALE G_SCALE
#endif
#if RGB_RED == 2
#define C2_SCALE R_SCALE
#endif
#if RGB_BLUE == 2
#define C2_SCALE B_SCALE
#endif


/*
 * First we have the histogram data structure and routines for creating it.
 *
 * The number of bits of precision can be adjusted by changing these symbols.
 * We recommend keeping 6 bits for G and 5 each for R and B.
 * If you have plenty of memory and cycles, 6 bits all around gives marginally
 * better results; if you are short of memory, 5 bits all around will save
 * some space but degrade the results.
 * To maintain a fully accurate histogram, we'd need to allocate a "long"
 * (preferably unsigned long) for each cell.  In practice this is overkill;
 * we can get by with 16 bits per cell.  Few of the cell counts will overflow,
 * and clamping those that do overflow to the maximum value will give close-
 * enough results.  This reduces the recommended histogram size from 256Kb
 * to 128Kb, which is a useful savings on PC-class machines.
 * (In the second pass the histogram space is re-used for pixel mapping data;
 * in that capacity, each cell must be able to store zero to the number of
 * desired colors.  16 bits/cell is plenty for that too.)
 * Since the JPEG code is intended to run in small memory model on 80x86
 * machines, we can't just allocate the histogram in one chunk.  Instead
 * of a true 3-D array, we use a row of pointers to 2-D arrays.  Each
 * pointer corresponds to a C0 value (typically 2^5 = 32 pointers) and
 * each 2-D array has 2^6*2^5 = 2048 or 2^6*2^6 = 4096 entries.  Note that
 * on 80x86 machines, the pointer row is in near memory but the actual
 * arrays are in far memory (same arrangement as we use for image arrays).
 */

#define MAXNUMCOLORS  (MAXJSAMPLE+1) /* maximum size of colormap */

/* These will do the right thing for either R,G,B or B,G,R color order,
 * but you may not like the results for other color orders.
 */
#define HIST_C0_BITS  5         /* bits of precision in R/B histogram */
#define HIST_C1_BITS  6         /* bits of precision in G histogram */
#define HIST_C2_BITS  5         /* bits of precision in B/R histogram */

/* Number of elements along histogram axes. */
#define HIST_C0_ELEMS  (1<<HIST_C0_BITS)
#define HIST_C1_ELEMS  (1<<HIST_C1_BITS)
#define HIST_C2_ELEMS  (1<<HIST_C2_BITS)

/* These are the amounts to shift an input value to get a histogram index. */
#define C0_SHIFT  (BITS_IN_JSAMPLE-HIST_C0_BITS)
#define C1_SHIFT  (BITS_IN_JSAMPLE-HIST_C1_BITS)
#define C2_SHIFT  (BITS_IN_JSAMPLE-HIST_C2_BITS)


typedef UINT16 histcell;        /* histogram cell; prefer an unsigned type */

typedef histcell FAR * histptr; /* for pointers to histogram cells */

typedef histcell hist1d[HIST_C2_ELEMS]; /* typedefs for the array */
typedef hist1d FAR * hist2d;    /* type for the 2nd-level pointers */
typedef hist2d * hist3d;        /* type for top-level pointer */


/* Declarations for Floyd-Steinberg dithering.
 *
 * Errors are accumulated into the array fserrors[], at a resolution of
 * 1/16th of a pixel count.  The error at a given pixel is propagated
 * to its not-yet-processed neighbors using the standard F-S fractions,
 *              ...     (here)  7/16
 *              3/16    5/16    1/16
 * We work left-to-right on even rows, right-to-left on odd rows.
 *
 * We can get away with a single array (holding one row's worth of errors)
 * by using it to store the current row's errors at pixel columns not yet
 * processed, but the next row's errors at columns already processed.  We
 * need only a few extra variables to hold the errors immediately around the
 * current column.  (If we are lucky, those variables are in registers, but
 * even if not, they're probably cheaper to access than array elements are.)
 *
 * The fserrors[] array has (#columns + 2) entries; the extra entry at
 * each end saves us from special-casing the first and last pixels.
 * Each entry is three values long, one value for each color component.
 *
 * Note: on a wide image, we might not have enough room in a PC's near data
 * segment to hold the error array; so it is allocated with alloc_large.
 */

#if BITS_IN_JSAMPLE == 8
typedef INT16 FSERROR;          /* 16 bits should be enough */
typedef int LOCFSERROR;         /* use 'int' for calculation temps */
#else
typedef INT32 FSERROR;          /* may need more than 16 bits */
typedef INT32 LOCFSERROR;       /* be sure calculation temps are big enough */
#endif

typedef FSERROR FAR *FSERRPTR;  /* pointer to error array (in FAR storage!) */


/* Private subobject */

typedef struct {
  struct jpeg_color_quantizer pub; /* public fields */

  /* Space for the eventually created colormap is stashed here */
  JSAMPARRAY sv_colormap;       /* colormap allocated at init time */
  int desired;                  /* desired # of colors = size of colormap */

  /* Variables for accumulating image statistics */
  hist3d histogram;             /* pointer to the histogram */

  boolean needs_zeroed;         /* TRUE if next pass must zero histogram */

  /* Variables for Floyd-Steinberg dithering */
  FSERRPTR fserrors;            /* accumulated errors */
  boolean on_odd_row;           /* flag to remember which row we are on */
  int * error_limiter;          /* table for clamping the applied error */
} my_cquantizer;

typedef my_cquantizer * my_cquantize_ptr;


/*
 * Prescan some rows of pixels.
 * In this module the prescan simply updates the histogram, which has been
 * initialized to zeroes by start_pass.
 * An output_buf parameter is required by the method signature, but no data
 * is actually output (in fact the buffer controller is probably passing a
 * NULL pointer).
 */

METHODDEF(void)
prescan_quantize (j_decompress_ptr cinfo, JSAMPARRAY input_buf,
                  JSAMPARRAY output_buf, int num_rows)
{
  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
  register JSAMPROW ptr;
  register histptr histp;
  register hist3d histogram = cquantize->histogram;
  int row;
  JDIMENSION col;
  JDIMENSION width = cinfo->output_width;

  for (row = 0; row < num_rows; row++) {
    ptr = input_buf[row];
    for (col = width; col > 0; col--) {
      /* get pixel value and index into the histogram */
      histp = & histogram[GETJSAMPLE(ptr[0]) >> C0_SHIFT]
                         [GETJSAMPLE(ptr[1]) >> C1_SHIFT]
                         [GETJSAMPLE(ptr[2]) >> C2_SHIFT];
      /* increment, check for overflow and undo increment if so. */
      if (++(*histp) <= 0)
        (*histp)--;
      ptr += 3;
    }
  }
}


/*
 * Next we have the really interesting routines: selection of a colormap
 * given the completed histogram.
 * These routines work with a list of "boxes", each representing a rectangular
 * subset of the input color space (to histogram precision).
 */

typedef struct {
  /* The bounds of the box (inclusive); expressed as histogram indexes */
  int c0min, c0max;
  int c1min, c1max;
  int c2min, c2max;
  /* The volume (actually 2-norm) of the box */
  INT32 volume;
  /* The number of nonzero histogram cells within this box */
  long colorcount;
} box;

typedef box * boxptr;


LOCAL(boxptr)
find_biggest_color_pop (boxptr boxlist, int numboxes)
/* Find the splittable box with the largest color population */
/* Returns NULL if no splittable boxes remain */
{
  register boxptr boxp;
  register int i;
  register long maxc = 0;
  boxptr which = NULL;
  
  for (i = 0, boxp = boxlist; i < numboxes; i++, boxp++) {
    if (boxp->colorcount > maxc && boxp->volume > 0) {
      which = boxp;
      maxc = boxp->colorcount;
    }
  }
  return which;
}


LOCAL(boxptr)
find_biggest_volume (boxptr boxlist, int numboxes)
/* Find the splittable box with the largest (scaled) volume */
/* Returns NULL if no splittable boxes remain */
{
  register boxptr boxp;
  register int i;
  register INT32 maxv = 0;
  boxptr which = NULL;
  
  for (i = 0, boxp = boxlist; i < numboxes; i++, boxp++) {
    if (boxp->volume > maxv) {
      which = boxp;
      maxv = boxp->volume;
    }
  }
  return which;
}


LOCAL(void)
update_box (j_decompress_ptr cinfo, boxptr boxp)
/* Shrink the min/max bounds of a box to enclose only nonzero elements, */
/* and recompute its volume and population */
{
  my_cquantize_ptr cquantize = (my_cquantize_ptr) cinfo->cquantize;
  hist3d histogram = cquantize->histogram;
  histptr histp;
  int c0,c1,c2;
  int c0min,c0max,c1min,c1max,c2min,c2max;
  INT32 dist0,dist1,dist2;
  long ccount;
  
  c0min = boxp->c0min;  c0max = boxp->c0max;
  c1min = boxp->c1min;  c1max = boxp->c1max;
  c2min = boxp->c2min;  c2max = boxp->c2max;
  
  if (c0max > c0min)
    for (c0 = c0min; c0 <= c0max; c0++)
      for (c1 = c1min; c1 <= c1max; c1++) {
        histp = & histogram[c0][c1][c2min];
        for (c2 = c2min; c2 <= c2max; c2++)
          if (*histp++ != 0) {
            boxp->c0min = c0min = c0;
            goto have_c0min;
          }
      }
 have_c0min:
  if (c0max > c0min)
    for (c0 = c0max; c0 >= c0min; c0--)
      for (c1 = c1min; c1 <= c1max; c1++) {
        histp = & histogram[c0][c1][c2min];
        for (c2 = c2min; c2 <= c2max; c2++)
          if (*histp++ != 0) {
            boxp->c0max = c0max = c0;
            goto have_c0max;
          }
      }
 have_c0max:
  if (c1max > c1min)
    for (c1 = c1min; c1 <= c1max; c1++)
      for (c0 = c0min; c0 <= c0max; c0++) {
        histp = & histogram[c0][c1][c2min];
        for (c2 = c2min; c2 <= c2max; c2++)
          if (*histp++ != 0) {
            boxp->c1min = c1min = c1;
            goto have_c1min;
          }
      }
 have_c1min:
  if (c1max > c1min)
    for (c1 = c1max; c1 >= c1min; c1--)
      for (c0 = c0min; c0 <= c0max; c0++) {
        histp = & histogram[c0][c1][c2min];
        for (c2 = c2min; c2 <= c2max; c2++)
          if (*histp++ != 0) {
            boxp->c1max = c1max = c1;
            goto have_c1max;
          }
      }
 have_c1max:
  if (c2max > c2min)
    for (c2 = c2min; c2 <= c2max; c2++)
      for (c0 = c0min; c0 <= c0max; c0++) {
        histp = & histogram[c0][c1min][c2];
        for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
          if (*histp != 0) {
            boxp->c2min = c2min = c2;
            goto have_c2min;
          }
      }
 have_c2min:
  if (c2max > c2min)
    for (c2 = c2max; c2 >= c2min; c2--)
      for (c0 = c0min; c0 <= c0max; c0++) {
        histp = & histogram[c0][c1min][c2];
        for (c1 = c1min; c1 <= c1max; c1++, histp += HIST_C2_ELEMS)
          if (*histp != 0) {
            boxp->c2max = c2max = c2;
            goto have_c2max;
          }
      }
 have_c2max:

  /* Update box volume.
   * We use 2-norm rather than real volume here; this biases the method
   * against making long narrow boxes, and it has the side benefit that
   * a box is splittable iff norm > 0.
   * Since the differences are expressed in histogram-cell units,
   * we have to shift back to JSAMPLE units to get consistent distances;
   * after which, we scale according to the selected distance scale factors.
   */
  dist0 = ((c0max - c0min) << C0_SHIFT) * C0_SCALE;
  dist1 = ((c1max - c1min) << C1_SHIFT) * C1_SCALE;
  dist2 = ((c2max - c2min) << C2_SHIFT) * C2_SCALE;
  boxp->volume = dist0*dist0 + dist1*dist1 + dist2*dist2;
  
  /* Now scan remaining volume of box and compute population */
  ccount = 0;
  for (c0 = c0min; c0 <= c0max; c0++)
    for (c1 = c1min; c1 <= c1max; c1++) {
      histp = & histogram[c0][c1][c2min];
      for (c2 = c2min; c2 <= c2max; c2++, histp++)
        if (*histp != 0) {
          ccount++;
        }
    }
  boxp->colorcount = ccount;
}


LOCAL(int)
median_cut (j_decompress_ptr cinfo, boxptr boxlist, int numboxes,
            int desired_colors)
/* Repeatedly select and split the largest box until we have enough boxes */
{
  int n,lb;
  int c0,c1,c2,cmax;
  register boxptr b1,b2;

  while (numboxes < desired_colors) {
    /* Select box to split.
     * Current algorithm: by population for first half, then by volume.
     */
    if (numboxes*2 <= desired_colors) {
      b1 = find_biggest_color_pop(boxlist, numboxes);
    } else {
      b1 = find_biggest_volume(boxlist, numboxes);
    }
    if (b1 == NULL)             /* no splittable boxes left! */
      break;
    b2 = &boxlist[numboxes];    /* where new box will go */
    /* Copy the color bounds to the new box. */
    b2->c0max = b1->c0max; b2->c1max = b1->c1max; b2->c2max = b1->c2max;
    b2->c0min = b1->c0min; b2->c1min = b1->c1min; b2->c2min = b1->c2min;
    /* Choose which axis to split the box on.
     * Current algorithm: longest scaled axis.
     * See notes in update_box about scaling distances.
     */
    c0 = ((b1->c0max - b1->c0min) << C0_SHIFT) * C0_SCALE;
    c1 = ((b1->c1max - b1->c1min) << C1_SHIFT) * C1_SCALE;
    c2 = ((b1->c2max - b1->c2min) << C2_SHIFT) * C2_SCALE;
    /* We want to break any ties in favor of green, then red, blue last.
     * This code does the right thing for R,G,B or B,G,R color orders only.
     */
#if RGB_RED == 0
    cmax = c1; n = 1;
    if (c0 > cmax) { cmax = c0; n = 0; }
    if (c2 > cmax) { n = 2; }
#else
    cmax = c1; n = 1;
    if (c2 > cmax) { cmax = c2; n = 2; }
    if (c0 > cmax) { n = 0; }
#endif
    /* Choose split point along selected axis, and update box bounds.
     * Current algorithm: split at halfway point.
     * (Since the box has been shrunk to minimum volume,
     * any split will produce two nonempty subboxes.)
     * Note that lb value is max for lower box, so must be < old max.
     */
    switch (n) {
    case 0:
      lb = (b1->c0max + b1->c0min) / 2;
      b1->c0max = lb;
      b2->c0min = lb+1;
      break;
    case 1:
      lb = (b1->c1max + b1->c1min) / 2;
      b1->c1max = lb;
      b2->c1min = lb+1;
      break;
    case 2:
      lb = (b1->c2max + b1->c2min) / 2;
      b1->c2max = lb;
      b2->c2min = lb+1;
      break;
    }
    /* Update stats for boxes */
    update_box(cinfo, b1);