git.sesse.net Git - x264/blob - common/csp.c

   1 /*****************************************************************************
   2  * csp.c: h264 encoder library
   3  *****************************************************************************
   4  * Copyright (C) 2004 Laurent Aimar
   5  * $Id: csp.c,v 1.1 2004/06/03 19:27:06 fenrir Exp $
   6  *
   7  * Authors: Laurent Aimar <fenrir@via.ecp.fr>
   8  *
   9  * This program is free software; you can redistribute it and/or modify
  10  * it under the terms of the GNU General Public License as published by
  11  * the Free Software Foundation; either version 2 of the License, or
  12  * (at your option) any later version.
  13  *
  14  * This program is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  17  * GNU General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU General Public License
  20  * along with this program; if not, write to the Free Software
  21  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
  22  *****************************************************************************/
  23
  24 #include <stdlib.h>
  25 #include <stdio.h>
  26 #include <string.h>
  27
  28 #include "common.h"
  29
  30 static inline void plane_copy( uint8_t *dst, int i_dst,
  31                                uint8_t *src, int i_src, int w, int h)
  32 {
  33     for( ; h > 0; h-- )
  34     {
  35         memcpy( dst, src, w );
  36         dst += i_dst;
  37         src += i_src;
  38     }
  39 }
  40 static inline void plane_copy_vflip( uint8_t *dst, int i_dst,
  41                                      uint8_t *src, int i_src, int w, int h)
  42 {
  43     plane_copy( dst, i_dst, src + (h -1)*i_src, -i_src, w, h );
  44 }
  45
  46 static inline void plane_subsamplev2( uint8_t *dst, int i_dst,
  47                                       uint8_t *src, int i_src, int w, int h)
  48 {
  49     for( ; h > 0; h-- )
  50     {
  51         uint8_t *d = dst;
  52         uint8_t *s = src;
  53         int     i;
  54         for( i = 0; i < w; i++ )
  55         {
  56             *d++ = ( s[0] + s[i_src] + 1 ) >> 1;
  57             s++;
  58         }
  59         dst += i_dst;
  60         src += 2 * i_src;
  61     }
  62 }
  63
  64 static inline void plane_subsamplev2_vlip( uint8_t *dst, int i_dst,
  65                                            uint8_t *src, int i_src, int w, int h)
  66 {
  67     plane_subsamplev2( dst, i_dst, src + (2*h-1)*i_src, -i_src, w, h );
  68 }
  69
  70 static inline void plane_subsamplehv2( uint8_t *dst, int i_dst,
  71                                        uint8_t *src, int i_src, int w, int h)
  72 {
  73     for( ; h > 0; h-- )
  74     {
  75         uint8_t *d = dst;
  76         uint8_t *s = src;
  77         int     i;
  78         for( i = 0; i < w; i++ )
  79         {
  80             *d++ = ( s[0] + s[1] + s[i_src] + s[i_src+1] + 1 ) >> 2;
  81             s += 2;
  82         }
  83         dst += i_dst;
  84         src += 2 * i_src;
  85     }
  86 }
  87
  88 static inline void plane_subsamplehv2_vlip( uint8_t *dst, int i_dst,
  89                                             uint8_t *src, int i_src, int w, int h)
  90 {
  91     plane_subsamplehv2( dst, i_dst, src + (2*h-1)*i_src, -i_src, w, h );
  92 }
  93
  94 static void i420_to_i420( x264_frame_t *frm, x264_image_t *img,
  95                           int i_width, int i_height )
  96 {
  97     if( img->i_csp & X264_CSP_VFLIP )
  98     {
  99         plane_copy_vflip( frm->plane[0], frm->i_stride[0],
 100                           img->plane[0], img->i_stride[0],
 101                           i_width, i_height );
 102         plane_copy_vflip( frm->plane[1], frm->i_stride[1],
 103                           img->plane[1], img->i_stride[1],
 104                           i_width / 2, i_height / 2 );
 105         plane_copy_vflip( frm->plane[2], frm->i_stride[2],
 106                           img->plane[2], img->i_stride[2],
 107                           i_width / 2, i_height / 2 );
 108     }
 109     else
 110     {
 111         plane_copy( frm->plane[0], frm->i_stride[0],
 112                     img->plane[0], img->i_stride[0],
 113                     i_width, i_height );
 114         plane_copy( frm->plane[1], frm->i_stride[1],
 115                     img->plane[1], img->i_stride[1],
 116                     i_width / 2, i_height / 2 );
 117         plane_copy( frm->plane[2], frm->i_stride[2],
 118                     img->plane[2], img->i_stride[2],
 119                     i_width / 2, i_height / 2 );
 120     }
 121 }
 122
 123 static void yv12_to_i420( x264_frame_t *frm, x264_image_t *img,
 124                           int i_width, int i_height )
 125 {
 126     if( img->i_csp & X264_CSP_VFLIP )
 127     {
 128         plane_copy_vflip( frm->plane[0], frm->i_stride[0],
 129                           img->plane[0], img->i_stride[0],
 130                           i_width, i_height );
 131         plane_copy_vflip( frm->plane[2], frm->i_stride[2],
 132                           img->plane[1], img->i_stride[1],
 133                           i_width / 2, i_height / 2 );
 134         plane_copy_vflip( frm->plane[1], frm->i_stride[1],
 135                           img->plane[2], img->i_stride[2],
 136                           i_width / 2, i_height / 2 );
 137     }
 138     else
 139     {
 140         plane_copy( frm->plane[0], frm->i_stride[0],
 141                     img->plane[0], img->i_stride[0],
 142                     i_width, i_height );
 143         plane_copy( frm->plane[2], frm->i_stride[2],
 144                     img->plane[1], img->i_stride[1],
 145                     i_width / 2, i_height / 2 );
 146         plane_copy( frm->plane[1], frm->i_stride[1],
 147                     img->plane[2], img->i_stride[2],
 148                     i_width / 2, i_height / 2 );
 149     }
 150 }
 151
 152 static void i422_to_i420( x264_frame_t *frm, x264_image_t *img,
 153                           int i_width, int i_height )
 154 {
 155     if( img->i_csp & X264_CSP_VFLIP )
 156     {
 157         plane_copy_vflip( frm->plane[0], frm->i_stride[0],
 158                           img->plane[0], img->i_stride[0],
 159                           i_width, i_height );
 160
 161         plane_subsamplev2_vlip( frm->plane[1], frm->i_stride[1],
 162                                 img->plane[1], img->i_stride[1],
 163                                 i_width / 2, i_height / 2 );
 164         plane_subsamplev2_vlip( frm->plane[2], frm->i_stride[2],
 165                                 img->plane[2], img->i_stride[2],
 166                                 i_width / 2, i_height / 2 );
 167     }
 168     else
 169     {
 170         plane_copy( frm->plane[0], frm->i_stride[0],
 171                     img->plane[0], img->i_stride[0],
 172                     i_width, i_height );
 173
 174         plane_subsamplev2( frm->plane[1], frm->i_stride[1],
 175                            img->plane[1], img->i_stride[1],
 176                            i_width / 2, i_height / 2 );
 177         plane_subsamplev2( frm->plane[2], frm->i_stride[2],
 178                            img->plane[2], img->i_stride[2],
 179                            i_width / 2, i_height / 2 );
 180     }
 181 }
 182
 183 static void i444_to_i420( x264_frame_t *frm, x264_image_t *img,
 184                           int i_width, int i_height )
 185 {
 186     if( img->i_csp & X264_CSP_VFLIP )
 187     {
 188         plane_copy_vflip( frm->plane[0], frm->i_stride[0],
 189                           img->plane[0], img->i_stride[0],
 190                           i_width, i_height );
 191
 192         plane_subsamplehv2_vlip( frm->plane[1], frm->i_stride[1],
 193                                  img->plane[1], img->i_stride[1],
 194                                  i_width / 2, i_height / 2 );
 195         plane_subsamplehv2_vlip( frm->plane[2], frm->i_stride[2],
 196                                  img->plane[2], img->i_stride[2],
 197                                  i_width / 2, i_height / 2 );
 198     }
 199     else
 200     {
 201         plane_copy( frm->plane[0], frm->i_stride[0],
 202                     img->plane[0], img->i_stride[0],
 203                     i_width, i_height );
 204
 205         plane_subsamplehv2( frm->plane[1], frm->i_stride[1],
 206                             img->plane[1], img->i_stride[1],
 207                             i_width / 2, i_height / 2 );
 208         plane_subsamplehv2( frm->plane[2], frm->i_stride[2],
 209                             img->plane[2], img->i_stride[2],
 210                             i_width / 2, i_height / 2 );
 211     }
 212 }
 213 static void yuyv_to_i420( x264_frame_t *frm, x264_image_t *img,
 214                           int i_width, int i_height )
 215 {
 216     uint8_t *src = img->plane[0];
 217     int     i_src= img->i_stride[0];
 218
 219     uint8_t *y   = frm->plane[0];
 220     uint8_t *u   = frm->plane[1];
 221     uint8_t *v   = frm->plane[2];
 222
 223     if( img->i_csp & X264_CSP_VFLIP )
 224     {
 225         src += ( i_height - 1 ) * i_src;
 226         i_src = -i_src;
 227     }
 228
 229     for( ; i_height > 0; i_height -= 2 )
 230     {
 231         uint8_t *ss = src;
 232         uint8_t *yy = y;
 233         uint8_t *uu = u;
 234         uint8_t *vv = v;
 235         int w;
 236
 237         for( w = i_width; w > 0; w -= 2 )
 238         {
 239             *yy++ = ss[0];
 240             *yy++ = ss[2];
 241
 242             *uu++ = ( ss[1] + ss[1+i_src] + 1 ) >> 1;
 243             *vv++ = ( ss[3] + ss[3+i_src] + 1 ) >> 1;
 244
 245             ss += 4;
 246         }
 247         src += i_src;
 248         y += frm->i_stride[0];
 249         u += frm->i_stride[1];
 250         v += frm->i_stride[2];
 251
 252         ss = src;
 253         yy = y;
 254         for( w = i_width; w > 0; w -= 2 )
 255         {
 256             *yy++ = ss[0];
 257             *yy++ = ss[2];
 258             ss += 4;
 259         }
 260         src += i_src;
 261         y += frm->i_stride[0];
 262     }
 263 }
 264
 265 /* Same value than in XviD */
 266 #define BITS 8
 267 #define FIX(f) ((int)((f) * (1 << BITS) + 0.5))
 268
 269 #define Y_R   FIX(0.257)
 270 #define Y_G   FIX(0.504)
 271 #define Y_B   FIX(0.098)
 272 #define Y_ADD 16
 273
 274 #define U_R   FIX(0.148)
 275 #define U_G   FIX(0.291)
 276 #define U_B   FIX(0.439)
 277 #define U_ADD 128
 278
 279 #define V_R   FIX(0.439)
 280 #define V_G   FIX(0.368)
 281 #define V_B   FIX(0.071)
 282 #define V_ADD 128
 283 #define RGB_TO_I420( name, POS_R, POS_G, POS_B, S_RGB ) \
 284 static void name( x264_frame_t *frm, x264_image_t *img, \
 285                   int i_width, int i_height )           \
 286 {                                                       \
 287     uint8_t *src = img->plane[0];                       \
 288     int     i_src= img->i_stride[0];                    \
 289     int     i_y  = frm->i_stride[0];                    \
 290     uint8_t *y   = frm->plane[0];                       \
 291     uint8_t *u   = frm->plane[1];                       \
 292     uint8_t *v   = frm->plane[2];                       \
 293                                                         \
 294     if( img->i_csp & X264_CSP_VFLIP )                   \
 295     {                                                   \
 296         src += ( i_height - 1 ) * i_src;                \
 297         i_src = -i_src;                                 \
 298     }                                                   \
 299                                                         \
 300     for(  ; i_height > 0; i_height -= 2 )               \
 301     {                                                   \
 302         uint8_t *ss = src;                              \
 303         uint8_t *yy = y;                                \
 304         uint8_t *uu = u;                                \
 305         uint8_t *vv = v;                                \
 306         int w;                                          \
 307                                                         \
 308         for( w = i_width; w > 0; w -= 2 )               \
 309         {                                               \
 310             int cr = 0,cg = 0,cb = 0;                   \
 311             int r, g, b;                                \
 312                                                         \
 313             /* Luma */                                  \
 314             cr = r = ss[POS_R];                         \
 315             cg = g = ss[POS_G];                         \
 316             cb = b = ss[POS_B];                         \
 317                                                         \
 318             yy[0] = Y_ADD + ((Y_R * r + Y_G * g + Y_B * b) >> BITS);    \
 319                                                         \
 320             cr+= r = ss[POS_R+i_src];                   \
 321             cg+= g = ss[POS_G+i_src];                   \
 322             cb+= b = ss[POS_B+i_src];                   \
 323             yy[i_y] = Y_ADD + ((Y_R * r + Y_G * g + Y_B * b) >> BITS);  \
 324             yy++;                                       \
 325             ss += S_RGB;                                \
 326                                                         \
 327             cr+= r = ss[POS_R];                         \
 328             cg+= g = ss[POS_G];                         \
 329             cb+= b = ss[POS_B];                         \
 330                                                         \
 331             yy[0] = Y_ADD + ((Y_R * r + Y_G * g + Y_B * b) >> BITS);    \
 332                                                         \
 333             cr+= r = ss[POS_R+i_src];                   \
 334             cg+= g = ss[POS_G+i_src];                   \
 335             cb+= b = ss[POS_B+i_src];                   \
 336             yy[i_y] = Y_ADD + ((Y_R * r + Y_G * g + Y_B * b) >> BITS);  \
 337             yy++;                                       \
 338             ss += S_RGB;                                \
 339                                                         \
 340             /* Chroma */                                \
 341             *uu++ = (uint8_t)(U_ADD + ((-U_R * cr - U_G * cg + U_B * cb) >> (BITS+2)) ); \
 342             *vv++ = (uint8_t)(V_ADD + (( V_R * cr - V_G * cg - V_B * cb) >> (BITS+2)) ); \
 343         }                                               \
 344                                                         \
 345         src += 2*i_src;                                   \
 346         y += 2*frm->i_stride[0];                        \
 347         u += frm->i_stride[1];                          \
 348         v += frm->i_stride[2];                          \
 349     }                                                   \
 350 }
 351
 352 RGB_TO_I420( rgb_to_i420,  0, 1, 2, 3 );
 353 RGB_TO_I420( bgr_to_i420,  2, 1, 0, 3 );
 354 RGB_TO_I420( bgra_to_i420, 2, 1, 0, 4 );
 355
 356 void x264_csp_init( int cpu, int i_csp, x264_csp_function_t *pf )
 357 {
 358     switch( i_csp )
 359     {
 360         case X264_CSP_I420:
 361             pf->i420 = i420_to_i420;
 362             pf->i422 = i422_to_i420;
 363             pf->i444 = i444_to_i420;
 364             pf->yv12 = yv12_to_i420;
 365             pf->yuyv = yuyv_to_i420;
 366             pf->rgb  = rgb_to_i420;
 367             pf->bgr  = bgr_to_i420;
 368             pf->bgra = bgra_to_i420;
 369             break;
 370
 371         default:
 372             /* For now, can't happen */
 373             fprintf( stderr, "arg in x264_csp_init\n" );
 374             exit( -1 );
 375             break;
 376     }
 377 }
 378