mpeg2enc/idct.c

   1 /* idct.c, inverse fast discrete cosine transform                           */
   2
   3 /* Copyright (C) 1996, MPEG Software Simulation Group. All Rights Reserved. */
   4
   5 /*
   6  * Disclaimer of Warranty
   7  *
   8  * These software programs are available to the user without any license fee or
   9  * royalty on an "as is" basis.  The MPEG Software Simulation Group disclaims
  10  * any and all warranties, whether express, implied, or statuary, including any
  11  * implied warranties or merchantability or of fitness for a particular
  12  * purpose.  In no event shall the copyright-holder be liable for any
  13  * incidental, punitive, or consequential damages of any kind whatsoever
  14  * arising from the use of these programs.
  15  *
  16  * This disclaimer of warranty extends to the user of these programs and user's
  17  * customers, employees, agents, transferees, successors, and assigns.
  18  *
  19  * The MPEG Software Simulation Group does not represent or warrant that the
  20  * programs furnished hereunder are free of infringement of any third-party
  21  * patents.
  22  *
  23  * Commercial implementations of MPEG-1 and MPEG-2 video, including shareware,
  24  * are subject to royalty fees to patent holders.  Many of these patents are
  25  * general enough such that they are unavoidable regardless of implementation
  26  * design.
  27  *
  28  */
  29
  30 /**********************************************************/
  31 /* inverse two dimensional DCT, Chen-Wang algorithm       */
  32 /* (cf. IEEE ASSP-32, pp. 803-816, Aug. 1984)             */
  33 /* 32-bit integer arithmetic (8 bit coefficients)         */
  34 /* 11 mults, 29 adds per DCT                              */
  35 /*                                      sE, 18.8.91       */
  36 /**********************************************************/
  37 /* coefficients extended to 12 bit for IEEE1180-1990      */
  38 /* compliance                           sE,  2.1.94       */
  39 /**********************************************************/
  40
  41 /* this code assumes >> to be a two's-complement arithmetic */
  42 /* right shift: (-2)>>1 == -1 , (-3)>>1 == -2               */
  43
  44 #include "config.h"
  45
  46 #define W1 2841 /* 2048*sqrt(2)*cos(1*pi/16) */
  47 #define W2 2676 /* 2048*sqrt(2)*cos(2*pi/16) */
  48 #define W3 2408 /* 2048*sqrt(2)*cos(3*pi/16) */
  49 #define W5 1609 /* 2048*sqrt(2)*cos(5*pi/16) */
  50 #define W6 1108 /* 2048*sqrt(2)*cos(6*pi/16) */
  51 #define W7 565  /* 2048*sqrt(2)*cos(7*pi/16) */
  52
  53 /* global declarations */
  54 void init_idct _ANSI_ARGS_((void));
  55 void idct _ANSI_ARGS_((short *block, unsigned char *temp));
  56
  57 /* private data */
  58 static short iclip[1024]; /* clipping table */
  59 static short *iclp;
  60
  61 /* private prototypes */
  62 static void idctrow _ANSI_ARGS_((short *blk));
  63 static void idctcol _ANSI_ARGS_((short *blk));
  64
  65 /* row (horizontal) IDCT
  66  *
  67  *           7                       pi         1
  68  * dst[k] = sum c[l] * src[l] * cos( -- * ( k + - ) * l )
  69  *          l=0                      8          2
  70  *
  71  * where: c[0]    = 128
  72  *        c[1..7] = 128*sqrt(2)
  73  */
  74
  75 static void idctrow(blk)
  76 short *blk;
  77 {
  78   int x0, x1, x2, x3, x4, x5, x6, x7, x8;
  79
  80   /* shortcut */
  81   if (!((x1 = blk[4]<<11) | (x2 = blk[6]) | (x3 = blk[2]) |
  82         (x4 = blk[1]) | (x5 = blk[7]) | (x6 = blk[5]) | (x7 = blk[3])))
  83   {
  84     blk[0]=blk[1]=blk[2]=blk[3]=blk[4]=blk[5]=blk[6]=blk[7]=blk[0]<<3;
  85     return;
  86   }
  87
  88   x0 = (blk[0]<<11) + 128; /* for proper rounding in the fourth stage */
  89
  90   /* first stage */
  91   x8 = W7*(x4+x5);
  92   x4 = x8 + (W1-W7)*x4;
  93   x5 = x8 - (W1+W7)*x5;
  94   x8 = W3*(x6+x7);
  95   x6 = x8 - (W3-W5)*x6;
  96   x7 = x8 - (W3+W5)*x7;
  97
  98   /* second stage */
  99   x8 = x0 + x1;
 100   x0 -= x1;
 101   x1 = W6*(x3+x2);
 102   x2 = x1 - (W2+W6)*x2;
 103   x3 = x1 + (W2-W6)*x3;
 104   x1 = x4 + x6;
 105   x4 -= x6;
 106   x6 = x5 + x7;
 107   x5 -= x7;
 108
 109   /* third stage */
 110   x7 = x8 + x3;
 111   x8 -= x3;
 112   x3 = x0 + x2;
 113   x0 -= x2;
 114   x2 = (181*(x4+x5)+128)>>8;
 115   x4 = (181*(x4-x5)+128)>>8;
 116
 117   /* fourth stage */
 118   blk[0] = (x7+x1)>>8;
 119   blk[1] = (x3+x2)>>8;
 120   blk[2] = (x0+x4)>>8;
 121   blk[3] = (x8+x6)>>8;
 122   blk[4] = (x8-x6)>>8;
 123   blk[5] = (x0-x4)>>8;
 124   blk[6] = (x3-x2)>>8;
 125   blk[7] = (x7-x1)>>8;
 126 }
 127
 128 /* column (vertical) IDCT
 129  *
 130  *             7                         pi         1
 131  * dst[8*k] = sum c[l] * src[8*l] * cos( -- * ( k + - ) * l )
 132  *            l=0                        8          2
 133  *
 134  * where: c[0]    = 1/1024
 135  *        c[1..7] = (1/1024)*sqrt(2)
 136  */
 137 static void idctcol(blk)
 138 short *blk;
 139 {
 140   int x0, x1, x2, x3, x4, x5, x6, x7, x8;
 141
 142   /* shortcut */
 143   if (!((x1 = (blk[8*4]<<8)) | (x2 = blk[8*6]) | (x3 = blk[8*2]) |
 144         (x4 = blk[8*1]) | (x5 = blk[8*7]) | (x6 = blk[8*5]) | (x7 = blk[8*3])))
 145   {
 146     blk[8*0]=blk[8*1]=blk[8*2]=blk[8*3]=blk[8*4]=blk[8*5]=blk[8*6]=blk[8*7]=
 147       iclp[(blk[8*0]+32)>>6];
 148     return;
 149   }
 150
 151   x0 = (blk[8*0]<<8) + 8192;
 152
 153   /* first stage */
 154   x8 = W7*(x4+x5) + 4;
 155   x4 = (x8+(W1-W7)*x4)>>3;
 156   x5 = (x8-(W1+W7)*x5)>>3;
 157   x8 = W3*(x6+x7) + 4;
 158   x6 = (x8-(W3-W5)*x6)>>3;
 159   x7 = (x8-(W3+W5)*x7)>>3;
 160
 161   /* second stage */
 162   x8 = x0 + x1;
 163   x0 -= x1;
 164   x1 = W6*(x3+x2) + 4;
 165   x2 = (x1-(W2+W6)*x2)>>3;
 166   x3 = (x1+(W2-W6)*x3)>>3;
 167   x1 = x4 + x6;
 168   x4 -= x6;
 169   x6 = x5 + x7;
 170   x5 -= x7;
 171
 172   /* third stage */
 173   x7 = x8 + x3;
 174   x8 -= x3;
 175   x3 = x0 + x2;
 176   x0 -= x2;
 177   x2 = (181*(x4+x5)+128)>>8;
 178   x4 = (181*(x4-x5)+128)>>8;
 179
 180   /* fourth stage */
 181   blk[8*0] = iclp[(x7+x1)>>14];
 182   blk[8*1] = iclp[(x3+x2)>>14];
 183   blk[8*2] = iclp[(x0+x4)>>14];
 184   blk[8*3] = iclp[(x8+x6)>>14];
 185   blk[8*4] = iclp[(x8-x6)>>14];
 186   blk[8*5] = iclp[(x0-x4)>>14];
 187   blk[8*6] = iclp[(x3-x2)>>14];
 188   blk[8*7] = iclp[(x7-x1)>>14];
 189 }
 190
 191 /* two dimensional inverse discrete cosine transform */
 192 void idct(short *block, unsigned char *temp)
 193 {
 194   int i;
 195
 196   for (i=0; i<8; i++)
 197     idctrow(block+8*i);
 198
 199   for (i=0; i<8; i++)
 200     idctcol(block+i);
 201 }
 202
 203 void init_idct()
 204 {
 205   int i;
 206
 207   iclp = iclip+512;
 208   for (i= -512; i<512; i++)
 209     iclp[i] = (i<-256) ? -256 : ((i>255) ? 255 : i);
 210 }