Create Subversion repository

git-svn-id: http://svn.code.sf.net/p/netpbm/code/trunk@1 9d0c8265-081b-0410-96cb-a4ca84ce46f8

1 files changed, 1045 insertions, 0 deletions

diff --git a/converter/ppm/ppmtompeg/block.c b/converter/ppm/ppmtompeg/block.c
new file mode 100644
index 00000000..913518c4
--- /dev/null
+++ b/converter/ppm/ppmtompeg/block.c
@@ -0,0 +1,1045 @@
+/*===========================================================================*
+ * block.c
+ *
+ *  Block routines
+ *
+ * NOTES:   MAD =   Mean Absolute Difference
+ *===========================================================================*/
+
+/* Copyright information is at end of file */
+
+#include <assert.h>
+
+#include "pm_c_util.h"
+#include "all.h"
+#include "mtypes.h"
+#include "frames.h"
+#include "bitio.h"
+#include "prototypes.h"
+#include "fsize.h"
+#include "opts.h"
+#include "postdct.h"
+
+#include "block.h"
+
+#define TRUNCATE_UINT8(x)   ((x < 0) ? 0 : ((x > 255) ? 255 : x))
+
+/*==================*
+ * GLOBAL VARIABLES *
+ *==================*/
+
+
+extern Block **dct, **dctb, **dctr;
+
+
+static vector
+halfVector(vector const v) {
+    vector half;
+
+    half.y = v.y/2;
+    half.x = v.x/2;
+
+    return half;
+}
+
+/*===========================*
+ * COMPUTE DCT OF DIFFERENCE *
+ *===========================*/
+
+/*===========================================================================*
+ *
+ *  compute current-motionBlock, take the DCT, and put the difference
+ *  back into current
+ *
+ * RETURNS: current block modified
+ *
+ * SIDE EFFECTS:    none
+ *
+ *===========================================================================*/
+void
+ComputeDiffDCTBlock(Block           current,
+                    Block           dest,
+                    Block           motionBlock,
+                    boolean * const significantDifferenceP) {
+
+    unsigned int y;
+    int diff;
+
+    diff = 0;  /* initial value */
+
+    for (y = 0; y < 8; ++y) {
+        unsigned int x;
+        for (x = 0; x < 8; ++x) {
+            current[y][x] -= motionBlock[y][x];
+            diff += ABS(current[y][x]);
+        }
+    }
+    /* Kill the block if change is too small     */
+    /* (block_bound defaults to 128, see opts.c) */
+    if (diff < block_bound)
+        *significantDifferenceP = FALSE;
+    else {
+        mp_fwd_dct_block2(current, dest);
+        *significantDifferenceP = TRUE;
+    }
+}
+
+
+
+/*===========================================================================*
+ *
+ *  appropriate (according to pattern, the coded block pattern) blocks
+ *  of 'current' are diff'ed and DCT'd.
+ *
+ * RETURNS: current blocks modified
+ *
+ * SIDE EFFECTS:    Can remove too-small difference blocks from pattern
+ *
+ * PRECONDITIONS:   appropriate blocks of 'current' have not yet been
+ *          modified
+ *
+ *===========================================================================*/
+void
+ComputeDiffDCTs(MpegFrame * const current,
+                MpegFrame * const prev,
+                int         const by,
+                int         const bx,
+                vector      const m,
+                int *       const patternP) {
+    
+    Block motionBlock;
+
+    if (collect_quant && (collect_quant_detailed & 1))
+        fprintf(collect_quant_fp, "l\n");
+    if (*patternP & 0x20) {
+        boolean significantDiff;
+        ComputeMotionBlock(prev->ref_y, by, bx, m, &motionBlock);
+        ComputeDiffDCTBlock(current->y_blocks[by][bx], dct[by][bx],
+                            motionBlock, &significantDiff);
+        if (!significantDiff)
+            *patternP ^= 0x20;
+    }
+
+    if (*patternP & 0x10) {
+        boolean significantDiff;
+        ComputeMotionBlock(prev->ref_y, by, bx+1, m, &motionBlock);
+        ComputeDiffDCTBlock(current->y_blocks[by][bx+1], dct[by][bx+1],
+                            motionBlock, &significantDiff);
+        if (!significantDiff)
+            *patternP ^= 0x10;
+    }
+
+    if (*patternP & 0x8) {
+        boolean significantDiff;
+        ComputeMotionBlock(prev->ref_y, by+1, bx, m, &motionBlock);
+        ComputeDiffDCTBlock(current->y_blocks[by+1][bx], dct[by+1][bx],
+                            motionBlock, &significantDiff);
+        if (!significantDiff)
+            *patternP ^= 0x8;
+    }
+
+    if (*patternP & 0x4) {
+        boolean significantDiff;
+        ComputeMotionBlock(prev->ref_y, by+1, bx+1, m, &motionBlock);
+        ComputeDiffDCTBlock(current->y_blocks[by+1][bx+1], dct[by+1][bx+1],
+                            motionBlock, &significantDiff);
+        if (!significantDiff)
+            *patternP ^= 0x4;
+    }
+
+    if (collect_quant && (collect_quant_detailed & 1))
+        fprintf(collect_quant_fp, "c\n");
+
+    if (*patternP & 0x2) {
+        boolean significantDiff;
+        ComputeMotionBlock(prev->ref_cb, by/2, bx/2, halfVector(m),
+                           &motionBlock);
+        ComputeDiffDCTBlock(current->cb_blocks[by/2][bx/2],
+                            dctb[by/2][bx/2], motionBlock,
+                            &significantDiff);
+        if (!significantDiff)
+            *patternP ^= 0x2;
+    }
+
+    if (*patternP & 0x1) {
+        boolean significantDiff;
+        ComputeMotionBlock(prev->ref_cr, by/2, bx/2, halfVector(m),
+                           &motionBlock);
+        ComputeDiffDCTBlock(current->cr_blocks[by/2][bx/2],
+                            dctr[by/2][bx/2], motionBlock,
+                            &significantDiff);
+        if (!significantDiff)
+            *patternP ^= 0x1;
+    }
+}
+
+
+
+static void
+computePrevFyFx(MpegFrame * const prevFrame,
+                int         const by,
+                int         const bx,
+                vector      const m,
+                uint8 ***   const prevP,
+                int *       const fyP,
+                int *       const fxP) {
+
+    boolean const xHalf = (ABS(m.x) % 2 == 1);
+    boolean const yHalf = (ABS(m.y) % 2 == 1);
+
+    MotionToFrameCoord(by, bx, m.y/2, m.x/2, fyP, fxP);
+
+    assert(*fyP >= 0); assert(*fxP >= 0);
+
+    /* C integer arithmetic rounds toward zero.  But what we need is a
+       "floor" -- i.e. round down.  So we adjust now for where the dividend
+       in the above divide by two was negative.
+    */
+
+    if (xHalf) {
+        if (m.x < 0)
+            --*fxP;
+
+        if (yHalf) {
+            if (m.y < 0)
+                --*fyP;
+        
+            *prevP = prevFrame->halfBoth;
+        } else
+            *prevP = prevFrame->halfX;
+    } else if (yHalf) {
+        if (m.y < 0)
+            --*fyP;
+        
+        *prevP = prevFrame->halfY;
+    } else
+        *prevP = prevFrame->ref_y;
+}
+
+
+
+/*======================*
+ * COMPUTE MOTION BLOCK *
+ *======================*/
+
+/*===========================================================================*
+ *
+ *  compute the motion-compensated block
+ *
+ * RETURNS: motionBlock
+ *
+ * SIDE EFFECTS:    none
+ *
+ * PRECONDITIONS:   motion vector MUST be valid
+ *
+ * NOTE:  could try to speed this up using halfX, halfY, halfBoth,
+ *    but then would have to compute for chrominance, and it's just
+ *    not worth the trouble (this procedure is not called relatively
+ *    often -- a constant number of times per macroblock)
+ *
+ *===========================================================================*/
+void
+ComputeMotionBlock(uint8 ** const prev,
+                   int      const by,
+                   int      const bx,
+                   vector   const m,
+                   Block *  const motionBlockP) {
+
+    int fy, fx;
+    boolean xHalf, yHalf;
+
+    xHalf = (ABS(m.x) % 2 == 1);
+    yHalf = (ABS(m.y) % 2 == 1);
+
+    MotionToFrameCoord(by, bx, m.y/2, m.x/2, &fy, &fx);
+
+    if (xHalf && yHalf) {
+        unsigned int y;
+        /* really should be fy+y-1 and fy+y so do (fy-1)+y = fy+y-1 and
+           (fy-1)+y+1 = fy+y
+        */
+        if (m.y < 0)
+            --fy;
+
+        if (m.x < 0)
+            --fx;
+    
+        for (y = 0; y < 8; ++y) {
+            int16 * const destPtr = (*motionBlockP)[y];
+            uint8 * const srcPtr  = &(prev[fy+y][fx]);
+            uint8 * const srcPtr2 = &(prev[fy+y+1][fx]);
+            unsigned int x;
+
+            for (x = 0; x < 8; ++x)
+                destPtr[x] =
+                    (srcPtr[x]+srcPtr[x+1]+srcPtr2[x]+srcPtr2[x+1]+2) >> 2;
+        }
+    } else if (xHalf) {
+        unsigned int y;
+        if (m.x < 0)
+            --fx;
+        
+        for (y = 0; y < 8; ++y) {
+            int16 * const destPtr = (*motionBlockP)[y];
+            uint8 * const srcPtr  = &(prev[fy+y][fx]);
+            unsigned int x;
+
+            for (x = 0; x < 8; ++x)
+                destPtr[x] = (srcPtr[x]+srcPtr[x+1]+1) >> 1;
+        }
+    } else if (yHalf) {
+        unsigned int y;
+        if ( m.y < 0 )
+            fy--;
+
+        for (y = 0; y < 8; ++y) {
+            int16 * const destPtr = (*motionBlockP)[y];
+            uint8 * const srcPtr  = &(prev[fy+y][fx]);
+            uint8 * const srcPtr2 = &(prev[fy+y+1][fx]);
+            unsigned int x;
+
+            for (x = 0; x < 8; ++x)
+                destPtr[x] = (srcPtr[x]+srcPtr2[x]+1) >> 1;
+        }
+    } else {
+        unsigned int y;
+        for (y = 0; y < 8; ++y) {
+            int16 * const destPtr = (*motionBlockP)[y];
+            uint8 * const srcPtr  = &(prev[fy+y][fx]);
+            unsigned int x;
+
+            for (x = 0; x < 8; ++x)
+                destPtr[x] = srcPtr[x];
+        }
+    }
+}
+
+
+
+/*===========================================================================*
+ *
+ *  compute the motion-compensated luminance block
+ *
+ * RETURNS: motionBlock
+ *
+ * SIDE EFFECTS:    none
+ *
+ * PRECONDITIONS:   motion vector MUST be valid
+ *
+ * NOTE:  see ComputeMotionBlock
+ *
+ *===========================================================================*/
+void
+ComputeMotionLumBlock(MpegFrame * const prevFrame,
+                      int         const by,
+                      int         const bx,
+                      vector      const m,
+                      LumBlock *  const motionBlockP) {
+
+    unsigned int y;
+    uint8 ** prev;
+    int fy, fx;
+
+    computePrevFyFx(prevFrame, by, bx, m, &prev, &fy, &fx);
+
+    for (y = 0; y < 16; ++y) {
+        uint8 * const across  = &(prev[fy+y][fx]);
+        int32 * const macross = motionBlockP->l[y];
+        unsigned int x;
+
+        for (x = 0; x < 16; ++x)
+            macross[x] = across[x];
+    }
+
+    /* this is what's really happening, in slow motion:
+     *
+     *  for (y = 0; y < 16; ++y, ++py)
+     *      for (x = 0; x < 16; ++x, ++px)
+     *          motionBlock[y][x] = prev[fy+y][fx+x];
+     *
+     */
+}
+
+
+/*=======================*
+ * BASIC ERROR FUNCTIONS *
+ *=======================*/
+
+
+/*===========================================================================*
+ *
+ *  return the MAD of two luminance blocks
+ *
+ * RETURNS: the MAD, if less than bestSoFar, or some number bigger if not
+ *
+ * SIDE EFFECTS:    none
+ *
+ *===========================================================================*/
+int32
+LumBlockMAD(const LumBlock * const currentBlockP,
+            const LumBlock * const motionBlockP,
+            int32            const bestSoFar) {
+
+    int32 diff;    /* max value of diff is 255*256 = 65280 */
+    unsigned int y;
+
+    diff = 0;  /* initial value */
+
+    for (y = 0; y < 16; ++y) {
+        const int32 * const currentRow = currentBlockP->l[y];
+        const int32 * const motionRow  = motionBlockP->l[y];
+        unsigned int x;
+        for (x = 0; x < 16; ++x)
+            diff += ABS(currentRow[x] - motionRow[x]);
+
+        if (diff > bestSoFar)
+            /* We already know the MAD won't be less than bestSoFar;
+               Caller doesn't care by how much we missed, so just return
+               this.
+            */
+            return diff;
+    }
+    /* Return the actual MAD */
+    return diff;
+}
+
+
+
+/*===========================================================================*
+ *
+ *  return the MAD of the currentBlock and the motion-compensated block
+ *      (without TUNEing)
+ *
+ *  (by, bx) is the location of the block in the frame 
+ *  (block number coordinates).  'm' is the motion of the block in pixels.
+ *  The moved block must be wholly within the frame.
+ *
+ * RETURNS: the MAD, if less than bestSoFar, or
+ *      some number bigger if not
+ *
+ * SIDE EFFECTS:    none
+ *
+ * PRECONDITIONS:  motion vector MUST be valid
+ *
+ * NOTES:  this is the procedure that is called the most, and should therefore
+ *         be the most optimized!!!
+ *
+ *===========================================================================*/
+int32
+LumMotionError(const LumBlock * const currentBlockP,
+               MpegFrame *      const prevFrame,
+               int              const by,
+               int              const bx,
+               vector           const m,
+               int32            const bestSoFar) {
+
+    int32 adiff;
+    int32 diff;    /* max value of diff is 255*256 = 65280 */
+    int y;
+    uint8 **prev;
+    int fy, fx;
+
+    computePrevFyFx(prevFrame, by, bx, m, &prev, &fy, &fx);
+
+    assert(fy >= 0); assert(fx >= 0);
+
+    adiff = 0;  /* initial value */
+    diff = 0; /* initial value */
+
+    switch (SearchCompareMode) {
+    case DEFAULT_SEARCH: /* Default. */
+        for (y = 0; y < 16; ++y) {
+            const int32 * const cacross = currentBlockP->l[y];
+            uint8 *       const across  = &prev[fy+y][fx];
+            unsigned int x;
+            
+            for (x = 0; x < 16; ++x) {
+                int32 const localDiff = across[x]-cacross[x];
+                diff += ABS(localDiff);
+            }
+            if (diff > bestSoFar)
+                return diff;
+        }
+        break;
+      
+    case LOCAL_DCT: {
+        Block     dctdiff[4], dctquant[4];
+        FlatBlock quant;
+        int x, i, tmp;
+        int distortion=0, datarate=0;
+        int pq = GetPQScale();
+      
+        for (y = 0;  y < 16;  ++y) {
+            const int32 * const cacross = currentBlockP->l[y];
+            uint8 * const across = &(prev[fy+y][fx]);
+            for (x = 0;  x < 16;  ++x) {
+                dctdiff[(x>7)+2*(y>7)][y%8][x%8] = cacross[x]-across[x];
+            }}
+
+        /* Calculate rate */
+        for (i = 0;  i < 4;  ++i) {
+            mp_fwd_dct_block2(dctdiff[i], dctdiff[i]);
+            if (Mpost_QuantZigBlock(dctdiff[i], quant, pq, FALSE) ==
+                MPOST_ZERO) {
+                /* no sense in continuing */
+                memset((char *)dctquant[i], 0, sizeof(Block));
+            } else {
+                Mpost_UnQuantZigBlock(quant, dctquant[i], pq, FALSE);
+                mpeg_jrevdct((int16 *)dctquant[i]);
+                datarate += CalcRLEHuffLength(quant);
+            }
+        }
+      
+        /* Calculate distortion */
+        for (y = 0;  y < 16;  ++y) {
+            const int32 * const cacross = currentBlockP->l[y];
+            uint8 * const across = &(prev[fy+y][fx]);
+            for (x = 0;  x < 16;  ++x) {
+                tmp = across[x] - cacross[x] +
+                    dctquant[(x>7)+2*(y>7)][y%8][x%8];
+                distortion += tmp*tmp;
+            }}
+        distortion /= 256;
+        distortion *= LocalDCTDistortScale;
+        datarate *= LocalDCTRateScale;
+        diff = (int) sqrt(distortion*distortion + datarate*datarate);
+        break;
+    }
+
+    case NO_DC_SEARCH: {
+        extern int32 niqtable[];
+        int pq = niqtable[0]*GetPQScale();
+        unsigned int y;
+        
+        for (y = 0; y < 16; ++y) {
+            const int32 * const cacross = currentBlockP->l[y];
+            uint8 * const across = &(prev[fy+y][fx]);
+            unsigned int x;
+            
+            for (x = 0; x < 16; ++x) {
+                int32 const localDiff = across[x]-cacross[x];
+                diff += localDiff;
+                adiff += ABS(localDiff);
+            }
+        }
+
+        diff /= 64*pq;  /* diff is now the DC difference (with QSCALE 1) */
+        adiff -= 64*pq*ABS(diff);
+        diff = adiff;
+    }
+    break;
+
+    case DO_Mean_Squared_Distortion:
+        for (y = 0; y < 16; ++y) {
+            const int32 * const cacross = currentBlockP->l[y];
+            uint8 * const across = &(prev[fy+y][fx]);
+            unsigned int x;
+
+            for (x = 0; x < 16; ++x) {
+                int32 const localDiff = across[x] - cacross[x];
+                diff += localDiff * localDiff;
+            }
+            if (diff > bestSoFar)
+                return diff;
+        }
+        break;
+    } /* End of Switch */
+
+    return diff;
+}
+
+
+
+/*===========================================================================*
+ *
+ *  return the MAD of the currentBlock and the average of the blockSoFar
+ *  and the motion-compensated block (this is used for B-frame searches)
+ *
+ * RETURNS: the MAD, if less than bestSoFar, or
+ *      some number bigger if not
+ *
+ * SIDE EFFECTS:    none
+ *
+ * PRECONDITIONS:  motion vector MUST be valid
+ *
+ *===========================================================================*/
+int32
+LumAddMotionError(const LumBlock * const currentBlockP,
+                  const LumBlock * const blockSoFarP,
+                  MpegFrame *      const prevFrame,
+                  int              const by,
+                  int              const bx,
+                  vector           const m,
+                  int32            const bestSoFar) {
+
+    int32 diff;    /* max value of diff is 255*256 = 65280 */
+    int y;
+    uint8 **prev;
+    int fy, fx;
+
+    computePrevFyFx(prevFrame, by, bx, m, &prev, &fy, &fx);
+
+    diff = 0; /* initial value */
+
+    /* do we add 1 before dividing by two?  Yes -- see MPEG-1 doc page 46 */
+
+    for (y = 0; y < 16; ++y) {
+        unsigned int x;
+        const uint8 * const across  = &prev[fy+y][fx];
+        const int32 * const bacross = blockSoFarP->l[y];
+        const int32 * const cacross = currentBlockP->l[y];
+
+        for (x = 0; x < 16; ++x) {
+            int32 const localDiff =
+                ((across[x] + bacross[x] + 1) / 2) - cacross[x];
+            diff += ABS(localDiff);
+        }
+
+        if (diff > bestSoFar)
+            return diff;
+    }
+    
+    /* This is what's happening:
+     *
+     *  ComputeMotionLumBlock(prevFrame, by, bx, my, mx, lumMotionBlock);
+     *
+     *  for (y = 0; y < 16; ++y)
+     *      for (x = 0; x < 16; ++x) {
+     *          localDiff = currentBlock[y][x] - lumMotionBlock[y][x];
+     *          diff += ABS(localDiff);
+     *      }
+     */
+
+    return diff;
+}
+
+
+
+/*===========================================================================*
+ *
+ *  adds the motion-compensated block to the given block
+ *
+ * RETURNS: block modified
+ *
+ * SIDE EFFECTS:    none
+ *
+ * PRECONDITIONS:  motion vector MUST be valid
+ *
+ *===========================================================================*/
+void
+AddMotionBlock(Block          block,
+               uint8 ** const prev,
+               int      const by,
+               int      const bx,
+               vector   const m) {
+
+    int     fy, fx;
+    boolean xHalf, yHalf;
+
+    xHalf = (ABS(m.x) % 2 == 1);
+    yHalf = (ABS(m.y) % 2 == 1);
+
+    MotionToFrameCoord(by, bx, (m.y/2), (m.x/2), &fy, &fx);
+
+    if (xHalf && yHalf) {
+        unsigned int y;
+        /* really should be fy+y-1 and fy+y so do (fy-1)+y = fy+y-1 and
+           (fy-1)+y+1 = fy+y
+        */
+        if (m.y < 0)
+            --fy;
+        if (m.x < 0)
+            --fx;
+
+        for (y = 0; y < 8; ++y) {
+            unsigned int x;
+            for (x = 0; x < 8; ++x)
+                block[y][x] += (prev[fy+y][fx+x]+prev[fy+y][fx+x+1]+
+                                prev[fy+y+1][fx+x]+prev[fy+y+1][fx+x+1]+2)>>2;
+        }
+    } else if (xHalf) {
+        unsigned int y;
+        if (m.x < 0)
+            --fx;
+
+        for (y = 0; y < 8; ++y) {
+            unsigned int x;
+            for (x = 0; x < 8; ++x) {
+                block[y][x] += (prev[fy+y][fx+x]+prev[fy+y][fx+x+1]+1)>>1;
+            }
+        }
+    } else if ( yHalf ) {
+        unsigned int y;
+        if (m.y < 0)
+            --fy;
+
+        for (y = 0; y < 8; ++y) {
+            unsigned int x;
+            for (x = 0; x < 8; ++x) {
+                block[y][x] += (prev[fy+y][fx+x]+prev[fy+y+1][fx+x]+1)>>1;
+            }
+        }
+    } else {
+        unsigned int y;
+        for (y = 0; y < 8; ++y) {
+            unsigned int x;
+            for (x = 0; x < 8; ++x) {
+                block[y][x] += (int16)prev[fy+y][fx+x];
+            }
+        }
+    }
+}
+
+
+/*===========================================================================*
+ *
+ *  adds the motion-compensated B-frame block to the given block
+ *
+ * RETURNS: block modified
+ *
+ * SIDE EFFECTS:    none
+ *
+ * PRECONDITIONS:  motion vectors MUST be valid
+ *
+ *===========================================================================*/
+void
+AddBMotionBlock(Block          block,
+                uint8 ** const prev,
+                uint8 ** const next,
+                int      const by,
+                int      const bx,
+                int      const mode,
+                motion   const motion) {
+
+    unsigned int y;
+    Block prevBlock, nextBlock;
+
+    switch (mode) {
+    case MOTION_FORWARD:
+        AddMotionBlock(block, prev, by, bx, motion.fwd);
+        break;
+    case MOTION_BACKWARD:
+        AddMotionBlock(block, next, by, bx, motion.bwd);
+        break;
+    default:
+        ComputeMotionBlock(prev, by, bx, motion.fwd, &prevBlock);
+        ComputeMotionBlock(next, by, bx, motion.bwd, &nextBlock);
+    }
+    for (y = 0; y < 8; ++y) {
+        unsigned int x;
+        for (x = 0; x < 8; ++x) {
+            block[y][x] += (prevBlock[y][x] + nextBlock[y][x] + 1) / 2;
+        }
+    }
+}
+
+
+/*===========================================================================*
+ *
+ *  copies the given block into the appropriate data area
+ *
+ * RETURNS: data modified
+ *
+ * SIDE EFFECTS:    none
+ *
+ *===========================================================================*/
+void
+BlockToData(uint8 ** const data,
+            Block          block,
+            int      const by,
+            int      const bx) {
+
+    int x, y;
+    int fy, fx;
+    int16    blockItem;
+
+    BLOCK_TO_FRAME_COORD(by, bx, fy, fx);
+
+    for ( y = 0; y < 8; y++ ) {
+    for ( x = 0; x < 8; x++ ) {
+        blockItem = block[y][x];
+        data[fy+y][fx+x] = TRUNCATE_UINT8(blockItem);
+    }
+    }
+}
+
+
+/*===========================================================================*
+ *
+ *  copies data into appropriate blocks
+ *
+ * RETURNS: mf modified
+ *
+ * SIDE EFFECTS:    none
+ *
+ * NOTES:  probably shouldn't be in this file
+ *
+ *===========================================================================*/
+void
+BlockifyFrame(MpegFrame * const frameP) {
+
+    int dctx, dcty;
+    int x, y;
+    int bx, by;
+    int fy, fx;
+    int16 *destPtr;
+    uint8 *srcPtr;
+    int16 *destPtr2;
+    uint8 *srcPtr2;
+    Block *blockPtr;
+    Block *blockPtr2;
+
+    dctx = Fsize_x / DCTSIZE;
+    dcty = Fsize_y / DCTSIZE;
+
+    /*
+     * copy y data into y_blocks
+     */
+    for (by = 0; by < dcty; by++) {
+    fy = by*DCTSIZE;
+    for (bx = 0; bx < dctx; bx++) {
+        fx = bx*DCTSIZE;
+        blockPtr = (Block *) &(frameP->y_blocks[by][bx][0][0]);
+        for (y = 0; y < DCTSIZE; y++) {
+        destPtr = &((*blockPtr)[y][0]);
+        srcPtr = &(frameP->orig_y[fy+y][fx]);
+        for (x = 0; x < DCTSIZE; x++) {
+            destPtr[x] = srcPtr[x];
+        }
+        }
+    }
+    }
+
+    /*
+     * copy cr/cb data into cr/cb_blocks
+     */
+    for (by = 0; by < (dcty >> 1); by++) {
+    fy = by*DCTSIZE;
+    for (bx = 0; bx < (dctx >> 1); bx++) {
+        fx = bx*DCTSIZE;
+        blockPtr = (Block *) &(frameP->cr_blocks[by][bx][0][0]);
+        blockPtr2 = (Block *) &(frameP->cb_blocks[by][bx][0][0]);
+        for (y = 0; y < DCTSIZE; y++) {
+        destPtr = &((*blockPtr)[y][0]);
+        srcPtr = &(frameP->orig_cr[fy+y][fx]);
+        destPtr2 = &((*blockPtr2)[y][0]);
+        srcPtr2 = &(frameP->orig_cb[fy+y][fx]);
+        for (x = 0; x < DCTSIZE; x++) {
+            destPtr[x] = srcPtr[x];
+            destPtr2[x] = srcPtr2[x];
+        }
+        }
+    }
+    }
+}
+
+
+/*===========================================================================*
+ *                                       *
+ * UNUSED PROCEDURES                                 *
+ *                                       *
+ *  The following procedures are all unused by the encoder           *
+ *                                       *
+ *  They are listed here for your convenience.  You might want to use    *
+ *  them if you experiment with different search techniques          *
+ *                                       *
+ *===========================================================================*/
+
+#ifdef UNUSED_PROCEDURES
+
+/* this procedure calculates the subsampled motion block (obviously)
+ *
+ * for speed, this procedure is probably not called anywhere (it is
+ * incorporated directly into LumDiffA, LumDiffB, etc.
+ *
+ * but leave it here anyway for clarity
+ *
+ * (startY, startX) = (0,0) for A....(0,1) for B...(1,0) for C...(1,1) for D
+ *  
+ */
+void
+ComputeSubSampledMotionLumBlock(MpegFrame * const prevFrame,
+                                int         const by,
+                                int         const bx,
+                                int         const my,
+                                int         const mx,
+                                LumBlock    const motionBlock,
+                                int         const startY,
+                                int         const startX) {
+
+    uint8 *across;
+    int32 *macross;
+    int32 *lastx;
+    int y;
+    uint8 **prev;
+    int    fy, fx;
+    boolean xHalf, yHalf;
+
+    xHalf = (ABS(mx) % 2 == 1);
+    yHalf = (ABS(my) % 2 == 1);
+
+    MotionToFrameCoord(by, bx, my/2, mx/2, &fy, &fx);
+
+    if ( xHalf ) {
+    if ( mx < 0 ) {
+        fx--;
+    }
+
+    if ( yHalf ) {
+        if ( my < 0 ) {
+        fy--;
+        }
+        
+        prev = prevFrame->halfBoth;
+    } else {
+        prev = prevFrame->halfX;
+    }
+    } else if ( yHalf ) {
+    if ( my < 0 ) {
+        fy--;
+    }
+
+    prev = prevFrame->halfY;
+    } else {
+    prev = prevFrame->ref_y;
+    }
+
+    for ( y = startY; y < 16; y += 2 ) {
+    across = &(prev[fy+y][fx+startX]);
+    macross = &(motionBlock[y][startX]);
+    lastx = &(motionBlock[y][16]);
+    while ( macross < lastx ) {
+        (*macross) = (*across);
+        across += 2;
+        macross += 2;
+    }
+    }
+
+    /* this is what's really going on in slow motion:
+     *
+     *  for ( y = startY; y < 16; y += 2 )
+     *      for ( x = startX; x < 16; x += 2 )
+     *      motionBlock[y][x] = prev[fy+y][fx+x];
+     *
+     */
+}
+
+
+/*===========================================================================*
+ *
+ *  return the MAD of the currentBlock and the motion-compensated block,
+ *  subsampled 4:1 with given starting coordinates (startY, startX)
+ *
+ * RETURNS: the MAD
+ *
+ * SIDE EFFECTS:    none
+ *
+ * PRECONDITIONS:  motion vector MUST be valid
+ *
+ * NOTES:  this procedure is never called.  Instead, see subsample.c.  This
+ *         procedure is provided only for possible use in extensions
+ *
+ *===========================================================================*/
+int32
+LumMotionErrorSubSampled(LumBlock    const currentBlock,
+                         MpegFrame * const prevFrame,
+                         int         const by,
+                         int         const bx,
+                         int         const my,
+                         int         const mx,
+                         int         const startY,
+                         int         const startX) {
+
+    int32 diff;     /* max value of diff is 255*256 = 65280 */
+    int32 localDiff;
+    int32 *cacross;
+    uint8 *macross;
+    int32 *lastx;
+    int y;
+    uint8 **prev;
+    int    fy, fx;
+    boolean xHalf, yHalf;
+
+    xHalf = (ABS(mx) % 2 == 1);
+    yHalf = (ABS(my) % 2 == 1);
+
+    motionToFrameCoord(by, bx, my/2, mx/2, &fy, &fx);
+
+    if ( xHalf ) {
+    if ( mx < 0 ) {
+        fx--;
+    }
+
+    if ( yHalf ) {
+        if ( my < 0 ) {
+        fy--;
+        }
+        
+        prev = prevFrame->halfBoth;
+    } else {
+        prev = prevFrame->halfX;
+    }
+    } else if ( yHalf ) {
+    if ( my < 0 ) {
+        fy--;
+    }
+
+    prev = prevFrame->halfY;
+    } else {
+    prev = prevFrame->ref_y;
+    }
+
+    diff = 0; /* initial value */
+
+    for ( y = startY; y < 16; y += 2 ) {
+    macross = &(prev[fy+y][fx+startX]);
+    cacross = &(currentBlock[y][startX]);
+    lastx = &(currentBlock[y][16]);
+    while ( cacross < lastx ) {
+        localDiff = (*cacross)-(*macross);
+        diff += ABS(localDiff);
+        macross += 2;
+        cacross += 2;
+    }
+    }
+
+    /* this is what's really happening:
+     *
+     *  ComputeSubSampledMotionLumBlock(prevFrame, by, bx, my, mx,
+     *                  lumMotionBlock, startY, startX);
+     *
+     *  for ( y = startY; y < 16; y += 2 )
+     *      for ( x = startX; x < 16; x += 2 )
+     *      {
+     *          localDiff = currentBlock[y][x] - lumMotionBlock[y][x];
+     *      diff += ABS(localDiff);
+     *      }
+     *
+     */
+
+    return (int32)diff;
+}
+
+
+#endif /* UNUSED_PROCEDURES */
+/*
+ * Copyright (c) 1995 The Regents of the University of California.
+ * All rights reserved.
+ *
+ * Permission to use, copy, modify, and distribute this software and its
+ * documentation for any purpose, without fee, and without written agreement is
+ * hereby granted, provided that the above copyright notice and the following
+ * two paragraphs appear in all copies of this software.
+ *
+ * IN NO EVENT SHALL THE UNIVERSITY OF CALIFORNIA BE LIABLE TO ANY PARTY FOR
+ * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES ARISING OUT
+ * OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN IF THE UNIVERSITY OF
+ * CALIFORNIA HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ * THE UNIVERSITY OF CALIFORNIA SPECIFICALLY DISCLAIMS ANY WARRANTIES,
+ * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
+ * AND FITNESS FOR A PARTICULAR PURPOSE.  THE SOFTWARE PROVIDED HEREUNDER IS
+ * ON AN "AS IS" BASIS, AND THE UNIVERSITY OF CALIFORNIA HAS NO OBLIGATION TO
+ * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
+ */
+