dwt.c 44.1 KB
Newer Older
1 2
/*
 * Copyright (C) 2004-2010 Michael Niedermayer <michaelni@gmx.at>
3
 * Copyright (C) 2008 David Conrad
4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * FFmpeg is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include "libavutil/attributes.h"
#include "dsputil.h"
#include "dwt.h"
25
#include "libavcodec/x86/dwt.h"
26

27
void ff_slice_buffer_init(slice_buffer * buf, int line_count, int max_allocated_lines, int line_width, IDWTELEM * base_buffer)
28 29 30 31
{
    int i;

    buf->base_buffer = base_buffer;
32 33 34 35 36
    buf->line_count  = line_count;
    buf->line_width  = line_width;
    buf->data_count  = max_allocated_lines;
    buf->line        = av_mallocz (sizeof(IDWTELEM *) * line_count);
    buf->data_stack  = av_malloc (sizeof(IDWTELEM *) * max_allocated_lines);
37 38 39 40 41 42 43 44

    for(i = 0; i < max_allocated_lines; i++){
        buf->data_stack[i] = av_malloc (sizeof(IDWTELEM) * line_width);
    }

    buf->data_stack_top = max_allocated_lines - 1;
}

45
IDWTELEM * ff_slice_buffer_load_line(slice_buffer * buf, int line)
46 47 48 49 50 51 52 53 54 55 56 57 58 59
{
    IDWTELEM * buffer;

    assert(buf->data_stack_top >= 0);
    if (buf->line[line])
        return buf->line[line];

    buffer = buf->data_stack[buf->data_stack_top];
    buf->data_stack_top--;
    buf->line[line] = buffer;

    return buffer;
}

60
void ff_slice_buffer_release(slice_buffer * buf, int line)
61 62 63 64 65 66 67 68 69 70 71 72
{
    IDWTELEM * buffer;

    assert(line >= 0 && line < buf->line_count);
    assert(buf->line[line]);

    buffer = buf->line[line];
    buf->data_stack_top++;
    buf->data_stack[buf->data_stack_top] = buffer;
    buf->line[line] = NULL;
}

73
void ff_slice_buffer_flush(slice_buffer * buf)
74 75 76 77
{
    int i;
    for(i = 0; i < buf->line_count; i++){
        if (buf->line[i])
78
            ff_slice_buffer_release(buf, i);
79 80 81
    }
}

82
void ff_slice_buffer_destroy(slice_buffer * buf)
83 84
{
    int i;
85
    ff_slice_buffer_flush(buf);
86 87 88 89 90 91 92 93 94 95

    for(i = buf->data_count - 1; i >= 0; i--){
        av_freep(&buf->data_stack[i]);
    }
    av_freep(&buf->data_stack);
    av_freep(&buf->line);
}

static inline int mirror(int v, int m){
    while((unsigned)v > (unsigned)m){
96 97
        v = -v;
        if(v < 0) v+= 2*m;
98 99 100 101 102 103 104 105 106
    }
    return v;
}

static av_always_inline void
lift(DWTELEM *dst, DWTELEM *src, DWTELEM *ref,
     int dst_step, int src_step, int ref_step,
     int width, int mul, int add, int shift,
     int highpass, int inverse){
107 108 109
    const int mirror_left  = !highpass;
    const int mirror_right = (width & 1) ^ highpass;
    const int w            = (width >> 1) - 1 + (highpass & width);
110 111 112 113
    int i;

#define LIFT(src, ref, inv) ((src) + ((inv) ? - (ref) : + (ref)))
    if(mirror_left){
114 115 116
        dst[0] = LIFT(src[0], ((mul * 2 * ref[0] + add) >> shift), inverse);
        dst   += dst_step;
        src   += src_step;
117 118 119
    }

    for(i=0; i<w; i++){
120 121 122
        dst[i * dst_step] =
            LIFT(src[i * src_step],
                 ((mul * (ref[i * ref_step] + ref[(i + 1) * ref_step]) + add) >> shift),
123 124 125 126
                 inverse);
    }

    if(mirror_right){
127 128 129
        dst[w * dst_step] =
            LIFT(src[w * src_step],
                 ((mul * 2 * ref[w * ref_step] + add) >> shift),
130 131 132 133 134 135 136 137 138
                 inverse);
    }
}

static av_always_inline void
inv_lift(IDWTELEM *dst, IDWTELEM *src, IDWTELEM *ref,
         int dst_step, int src_step, int ref_step,
         int width, int mul, int add, int shift,
         int highpass, int inverse){
139 140 141
    const int mirror_left  = !highpass;
    const int mirror_right = (width&1) ^ highpass;
    const int w            = (width >> 1) - 1 + (highpass & width);
142 143 144 145
    int i;

#define LIFT(src, ref, inv) ((src) + ((inv) ? - (ref) : + (ref)))
    if(mirror_left){
146 147 148 149 150 151
        dst[0] =
            LIFT(src[0],
                 ((mul * 2 * ref[0] + add) >> shift),
                 inverse);
        dst   += dst_step;
        src   += src_step;
152 153
    }

154 155 156 157
    for(i = 0; i < w; i++){
        dst[i * dst_step] =
            LIFT(src[i * src_step],
                 ((mul * (ref[i * ref_step] + ref[(i + 1) * ref_step]) + add) >> shift),
158 159 160 161
                 inverse);
    }

    if(mirror_right){
162 163 164
        dst[w * dst_step] =
            LIFT(src[w * src_step],
                 ((mul * 2 * ref[w * ref_step] + add) >> shift),
165 166 167 168 169 170 171 172 173 174
                 inverse);
    }
}

#ifndef liftS
static av_always_inline void
liftS(DWTELEM *dst, DWTELEM *src, DWTELEM *ref,
      int dst_step, int src_step, int ref_step,
      int width, int mul, int add, int shift,
      int highpass, int inverse){
175 176 177
    const int mirror_left  = !highpass;
    const int mirror_right = (width&1) ^ highpass;
    const int w            = (width >> 1) - 1 + (highpass & width);
178 179 180
    int i;

    assert(shift == 4);
181 182 183 184
#define LIFTS(src, ref, inv)                                            \
    ((inv) ?                                                            \
     (src) + (((ref) + 4 * (src)) >> shift):                            \
     -((-16 * (src) + (ref) + add / 4 + 1 + (5 << 25)) / (5 * 4) - (1 << 23)))
185
    if(mirror_left){
186 187 188
        dst[0] = LIFTS(src[0], mul * 2 * ref[0] + add, inverse);
        dst   += dst_step;
        src   += src_step;
189 190
    }

191 192 193 194
    for(i = 0; i < w; i++){
        dst[i * dst_step] =
            LIFTS(src[i * src_step],
                  mul * (ref[i * ref_step] + ref[(i+1) * ref_step]) + add,
195 196 197 198
                  inverse);
    }

    if(mirror_right){
199 200
        dst[w * dst_step] =
            LIFTS(src[w * src_step], mul * 2 * ref[w * ref_step] + add, inverse);
201 202 203 204 205 206 207
    }
}
static av_always_inline void
inv_liftS(IDWTELEM *dst, IDWTELEM *src, IDWTELEM *ref,
          int dst_step, int src_step, int ref_step,
          int width, int mul, int add, int shift,
          int highpass, int inverse){
208 209 210
    const int mirror_left  = !highpass;
    const int mirror_right = (width&1) ^ highpass;
    const int w            = (width >> 1) - 1 + (highpass & width);
211 212 213
    int i;

    assert(shift == 4);
214 215 216 217
#define LIFTS(src, ref, inv)                                            \
    ((inv) ?                                                            \
     (src) + (((ref) + 4 * (src)) >> shift):                            \
     -((-16 * (src) + (ref) + add / 4 + 1 + (5 << 25)) / (5 * 4) - (1 << 23)))
218
    if(mirror_left){
219
        dst[0] = LIFTS(src[0], mul * 2 * ref[0] + add, inverse);
220 221 222 223
        dst += dst_step;
        src += src_step;
    }

224 225 226 227
    for(i = 0; i < w; i++){
        dst[i * dst_step] =
            LIFTS(src[i * src_step],
                  mul * (ref[i * ref_step] + ref[(i+1) * ref_step]) + add,
228 229 230 231
                  inverse);
    }

    if(mirror_right){
232 233
        dst[w * dst_step] =
            LIFTS(src[w * src_step], mul * 2 * ref[w * ref_step] + add, inverse);
234 235 236 237 238 239
    }
}
#endif /* ! liftS */

static void horizontal_decompose53i(DWTELEM *b, int width){
    DWTELEM temp[width];
240 241
    const int width2 = width>>1;
    const int w2     = (width+1)>>1;
242 243
    int x;

244 245 246
    for(x = 0; x < width2; x++){
        temp[x   ] = b[2 * x    ];
        temp[x+w2] = b[2 * x + 1];
247
    }
248 249
    if(width & 1)
        temp[x   ] = b[2 * x    ];
250 251 252
#if 0
    {
    int A1,A2,A3,A4;
253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275
    A2  = temp[1       ];
    A4  = temp[0       ];
    A1  = temp[0+width2];
    A1 -= (A2 + A4) >> 1;
    A4 += (A1 + 1) >> 1;
    b[0 + width2] = A1;
    b[0       ]   = A4;
    for(x = 1; x + 1 < width2; x += 2){
        A3  = temp[x + width2];
        A4  = temp[x + 1     ];
        A3 -= (A2 + A4) >> 1;
        A2 += (A1 + A3 + 2) >> 2;
        b[x + width2] = A3;
        b[x         ] = A2;

        A1  = temp[x + 1 + width2];
        A2  = temp[x + 2         ];
        A1 -= (A2 + A4) >> 1;
        A4 += (A1 + A3 + 2) >> 2;
        b[x + 1 + width2] = A1;
        b[x + 1         ] = A4;
    }
    A3  = temp[width - 1];
276
    A3 -= A2;
277
    A2 += (A1 + A3 + 2) >> 2;
278 279 280 281
    b[width -1] = A3;
    b[width2-1] = A2;
    }
#else
282 283
    lift(b + w2, temp + w2, temp, 1, 1, 1, width, -1, 0, 1, 1, 0);
    lift(b   , temp   , b + w2, 1, 1, 1, width,  1, 2, 2, 0, 0);
284 285 286 287 288 289
#endif /* 0 */
}

static void vertical_decompose53iH0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
    int i;

290 291
    for(i = 0; i < width; i++){
        b1[i] -= (b0[i] + b2[i]) >> 1;
292 293 294 295 296 297
    }
}

static void vertical_decompose53iL0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
    int i;

298 299
    for(i = 0; i < width; i++){
        b1[i] += (b0[i] + b2[i] + 2) >> 2;
300 301 302 303 304
    }
}

static void spatial_decompose53i(DWTELEM *buffer, int width, int height, int stride){
    int y;
305 306
    DWTELEM *b0 = buffer + mirror(-2 - 1, height-1)*stride;
    DWTELEM *b1 = buffer + mirror(-2    , height-1)*stride;
307

308 309 310
    for(y = -2; y < height; y += 2){
        DWTELEM *b2 = buffer + mirror(y + 1, height-1) * stride;
        DWTELEM *b3 = buffer + mirror(y + 2, height-1) * stride;
311

312 313
        if(y + 1 < (unsigned)height) horizontal_decompose53i(b2, width);
        if(y + 2 < (unsigned)height) horizontal_decompose53i(b3, width);
314

315 316
        if(y + 1 < (unsigned)height) vertical_decompose53iH0(b1, b2, b3, width);
        if(y + 0 < (unsigned)height) vertical_decompose53iL0(b0, b1, b2, width);
317

318 319
        b0 = b2;
        b1 = b3;
320 321 322 323 324
    }
}

static void horizontal_decompose97i(DWTELEM *b, int width){
    DWTELEM temp[width];
325
    const int w2 = (width+1)>>1;
326

327 328 329 330
    lift (temp + w2, b    + 1 , b        , 1, 2, 2, width,  W_AM, W_AO, W_AS, 1, 1);
    liftS(temp     , b        , temp + w2, 1, 2, 1, width,  W_BM, W_BO, W_BS, 0, 0);
    lift (b    + w2, temp + w2, temp     , 1, 1, 1, width,  W_CM, W_CO, W_CS, 1, 0);
    lift (b        , temp     , b    + w2, 1, 1, 1, width,  W_DM, W_DO, W_DS, 0, 0);
331 332 333 334 335 336
}


static void vertical_decompose97iH0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
    int i;

337 338
    for(i = 0; i < width; i++){
        b1[i] -= (W_AM * (b0[i] + b2[i]) + W_AO) >> W_AS;
339 340 341 342 343 344
    }
}

static void vertical_decompose97iH1(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
    int i;

345 346
    for(i=0; i < width; i++){
        b1[i] += (W_CM * (b0[i] + b2[i]) + W_CO) >> W_CS;
347 348 349 350 351 352
    }
}

static void vertical_decompose97iL0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
    int i;

353
    for(i = 0; i < width; i++){
354
#ifdef liftS
355
        b1[i] -= (W_BM * (b0[i] + b2[i]) + W_BO) >> W_BS;
356
#else
357
        b1[i] = (16 * 4 * b1[i] - 4 * (b0[i] + b2[i]) + W_BO * 5 + (5 << 27)) / (5 * 16) - (1 << 23);
358 359 360 361 362 363 364
#endif
    }
}

static void vertical_decompose97iL1(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){
    int i;

365 366
    for(i = 0; i < width; i++){
        b1[i] += (W_DM * (b0[i] + b2[i]) + W_DO) >> W_DS;
367 368 369 370 371
    }
}

static void spatial_decompose97i(DWTELEM *buffer, int width, int height, int stride){
    int y;
372 373 374 375
    DWTELEM *b0 = buffer + mirror(-4 - 1, height-1) * stride;
    DWTELEM *b1 = buffer + mirror(-4    , height-1) * stride;
    DWTELEM *b2 = buffer + mirror(-4 + 1, height-1) * stride;
    DWTELEM *b3 = buffer + mirror(-4 + 2, height-1) * stride;
376

377 378 379
    for(y = -4; y < height; y += 2){
        DWTELEM *b4 = buffer + mirror(y + 3, height-1) * stride;
        DWTELEM *b5 = buffer + mirror(y + 4, height-1) * stride;
380

381 382
        if(y + 3 < (unsigned)height) horizontal_decompose97i(b4, width);
        if(y + 4 < (unsigned)height) horizontal_decompose97i(b5, width);
383

384 385 386 387
        if(y + 3 < (unsigned)height) vertical_decompose97iH0(b3, b4, b5, width);
        if(y + 2 < (unsigned)height) vertical_decompose97iL0(b2, b3, b4, width);
        if(y + 1 < (unsigned)height) vertical_decompose97iH1(b1, b2, b3, width);
        if(y + 0 < (unsigned)height) vertical_decompose97iL1(b0, b1, b2, width);
388

389 390 391 392
        b0 = b2;
        b1 = b3;
        b2 = b4;
        b3 = b5;
393 394 395 396 397 398
    }
}

void ff_spatial_dwt(DWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count){
    int level;

399
    for(level = 0; level < decomposition_count; level++){
400
        switch(type){
401 402
        case DWT_97: spatial_decompose97i(buffer, width >> level, height >> level, stride << level); break;
        case DWT_53: spatial_decompose53i(buffer, width >> level, height >> level, stride << level); break;
403 404 405 406 407 408
        }
    }
}

static void horizontal_compose53i(IDWTELEM *b, int width){
    IDWTELEM temp[width];
409 410
    const int width2 = width       >> 1;
    const int w2     = (width + 1) >> 1;
411 412
    int x;

413 414 415
    for(x = 0; x < width2; x++){
        temp[2 * x    ] = b[x     ];
        temp[2 * x + 1] = b[x + w2];
416
    }
417 418
    if(width & 1)
        temp[2 * x    ] = b[x   ];
419

420 421 422 423
    b[0] = temp[0] - ((temp[1] + 1) >> 1);
    for(x = 2; x < width - 1; x += 2){
        b[x    ] = temp[x    ] - ((temp[x - 1] + temp[x + 1] + 2) >> 2);
        b[x - 1] = temp[x - 1] + ((b   [x - 2] + b   [x  ] + 1) >> 1);
424
    }
425 426 427
    if(width & 1){
        b[x    ] = temp[x    ] - ((temp[x - 1] + 1) >> 1);
        b[x - 1] = temp[x - 1] + ((b   [x - 2] + b  [x  ] + 1) >> 1);
428
    }else
429
        b[x - 1] = temp[x - 1] + b[x - 2];
430 431 432 433 434
}

static void vertical_compose53iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width){
    int i;

435 436
    for(i = 0; i < width; i++){
        b1[i] += (b0[i] + b2[i]) >> 1;
437 438 439 440 441 442
    }
}

static void vertical_compose53iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width){
    int i;

443 444
    for(i = 0; i < width; i++){
        b1[i] -= (b0[i] + b2[i] + 2) >> 2;
445 446 447 448
    }
}

static void spatial_compose53i_buffered_init(DWTCompose *cs, slice_buffer * sb, int height, int stride_line){
449 450 451
    cs->b0 = slice_buffer_get_line(sb, mirror(-1-1, height - 1) * stride_line);
    cs->b1 = slice_buffer_get_line(sb, mirror(-1  , height - 1) * stride_line);
    cs->y  = -1;
452 453 454
}

static void spatial_compose53i_init(DWTCompose *cs, IDWTELEM *buffer, int height, int stride){
455 456 457
    cs->b0 = buffer + mirror(-1-1, height - 1) * stride;
    cs->b1 = buffer + mirror(-1  , height - 1) * stride;
    cs->y  = -1;
458 459 460
}

static void spatial_compose53i_dy_buffered(DWTCompose *cs, slice_buffer * sb, int width, int height, int stride_line){
461 462 463 464 465
    int y        = cs->y;
    IDWTELEM *b0 = cs->b0;
    IDWTELEM *b1 = cs->b1;
    IDWTELEM *b2 = slice_buffer_get_line(sb, mirror(y + 1, height-1) * stride_line);
    IDWTELEM *b3 = slice_buffer_get_line(sb, mirror(y + 2, height-1) * stride_line);
466

467
    if(y + 1 < (unsigned)height && y < (unsigned)height){
468 469
        int x;

470 471 472
        for(x = 0; x < width; x++){
            b2[x] -= (b1[x] + b3[x] + 2) >> 2;
            b1[x] += (b0[x] + b2[x]) >> 1;
473 474
        }
    }else{
475 476
        if(y + 1 < (unsigned)height) vertical_compose53iL0(b1, b2, b3, width);
        if(y + 0 < (unsigned)height) vertical_compose53iH0(b0, b1, b2, width);
477 478
    }

479 480
        if(y - 1 <(unsigned)height) horizontal_compose53i(b0, width);
        if(y + 0 <(unsigned)height) horizontal_compose53i(b1, width);
481 482 483 484 485 486 487

    cs->b0 = b2;
    cs->b1 = b3;
    cs->y += 2;
}

static void spatial_compose53i_dy(DWTCompose *cs, IDWTELEM *buffer, int width, int height, int stride){
488 489 490 491 492
    int y        = cs->y;
    IDWTELEM *b0 = cs->b0;
    IDWTELEM *b1 = cs->b1;
    IDWTELEM *b2 = buffer + mirror(y + 1, height - 1) * stride;
    IDWTELEM *b3 = buffer + mirror(y + 2, height - 1) * stride;
493

494 495
        if(y + 1 < (unsigned)height) vertical_compose53iL0(b1, b2, b3, width);
        if(y + 0 < (unsigned)height) vertical_compose53iH0(b0, b1, b2, width);
496

497 498
        if(y - 1 < (unsigned)height) horizontal_compose53i(b0, width);
        if(y + 0 < (unsigned)height) horizontal_compose53i(b1, width);
499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514

    cs->b0 = b2;
    cs->b1 = b3;
    cs->y += 2;
}

static void av_unused spatial_compose53i(IDWTELEM *buffer, int width, int height, int stride){
    DWTCompose cs;
    spatial_compose53i_init(&cs, buffer, height, stride);
    while(cs.y <= height)
        spatial_compose53i_dy(&cs, buffer, width, height, stride);
}


void ff_snow_horizontal_compose97i(IDWTELEM *b, int width){
    IDWTELEM temp[width];
515
    const int w2 = (width + 1) >> 1;
516 517

#if 0 //maybe more understadable but slower
518 519
    inv_lift (temp    , b      , b  + w2, 2, 1, 1, width,  W_DM, W_DO, W_DS, 0, 1);
    inv_lift (temp + 1, b  + w2, temp   , 2, 1, 2, width,  W_CM, W_CO, W_CS, 1, 1);
520

521 522
    inv_liftS(b      ,temp    ,temp + 1, 2, 2, 2, width,  W_BM, W_BO, W_BS, 0, 1);
    inv_lift (b + 1  ,temp + 1,b       , 2, 2, 2, width,  W_AM, W_AO, W_AS, 1, 0);
523 524
#else
    int x;
525 526 527 528
    temp[0] = b[0] - ((3 * b[w2] + 2) >> 2);
    for(x = 1; x < (width >> 1); x++){
        temp[2 * x    ] = b[x         ] - (( 3 * (b   [ x + w2 - 1] + b[x + w2]) + 4) >> 3);
        temp[2 * x - 1] = b[x + w2 - 1] - temp[2 * x - 2] - temp[2 * x];
529
    }
530
    if(width & 1){
531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702
        temp[2*x  ] = b[x     ] - ((3*b   [x+w2-1]+2)>>2);
        temp[2*x-1] = b[x+w2-1] - temp[2*x-2] - temp[2*x];
    }else
        temp[2*x-1] = b[x+w2-1] - 2*temp[2*x-2];

    b[0] = temp[0] + ((2*temp[0] + temp[1]+4)>>3);
    for(x=2; x<width-1; x+=2){
        b[x  ] = temp[x  ] + ((4*temp[x  ] + temp[x-1] + temp[x+1]+8)>>4);
        b[x-1] = temp[x-1] + ((3*(b  [x-2] + b   [x  ] ))>>1);
    }
    if(width&1){
        b[x  ] = temp[x  ] + ((2*temp[x  ] + temp[x-1]+4)>>3);
        b[x-1] = temp[x-1] + ((3*(b  [x-2] + b   [x  ] ))>>1);
    }else
        b[x-1] = temp[x-1] + 3*b [x-2];
#endif
}

static void vertical_compose97iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width){
    int i;

    for(i=0; i<width; i++){
        b1[i] += (W_AM*(b0[i] + b2[i])+W_AO)>>W_AS;
    }
}

static void vertical_compose97iH1(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width){
    int i;

    for(i=0; i<width; i++){
        b1[i] -= (W_CM*(b0[i] + b2[i])+W_CO)>>W_CS;
    }
}

static void vertical_compose97iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width){
    int i;

    for(i=0; i<width; i++){
#ifdef liftS
        b1[i] += (W_BM*(b0[i] + b2[i])+W_BO)>>W_BS;
#else
        b1[i] += (W_BM*(b0[i] + b2[i])+4*b1[i]+W_BO)>>W_BS;
#endif
    }
}

static void vertical_compose97iL1(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width){
    int i;

    for(i=0; i<width; i++){
        b1[i] -= (W_DM*(b0[i] + b2[i])+W_DO)>>W_DS;
    }
}

void ff_snow_vertical_compose97i(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, IDWTELEM *b3, IDWTELEM *b4, IDWTELEM *b5, int width){
    int i;

    for(i=0; i<width; i++){
        b4[i] -= (W_DM*(b3[i] + b5[i])+W_DO)>>W_DS;
        b3[i] -= (W_CM*(b2[i] + b4[i])+W_CO)>>W_CS;
#ifdef liftS
        b2[i] += (W_BM*(b1[i] + b3[i])+W_BO)>>W_BS;
#else
        b2[i] += (W_BM*(b1[i] + b3[i])+4*b2[i]+W_BO)>>W_BS;
#endif
        b1[i] += (W_AM*(b0[i] + b2[i])+W_AO)>>W_AS;
    }
}

static void spatial_compose97i_buffered_init(DWTCompose *cs, slice_buffer * sb, int height, int stride_line){
    cs->b0 = slice_buffer_get_line(sb, mirror(-3-1, height-1) * stride_line);
    cs->b1 = slice_buffer_get_line(sb, mirror(-3  , height-1) * stride_line);
    cs->b2 = slice_buffer_get_line(sb, mirror(-3+1, height-1) * stride_line);
    cs->b3 = slice_buffer_get_line(sb, mirror(-3+2, height-1) * stride_line);
    cs->y = -3;
}

static void spatial_compose97i_init(DWTCompose *cs, IDWTELEM *buffer, int height, int stride){
    cs->b0 = buffer + mirror(-3-1, height-1)*stride;
    cs->b1 = buffer + mirror(-3  , height-1)*stride;
    cs->b2 = buffer + mirror(-3+1, height-1)*stride;
    cs->b3 = buffer + mirror(-3+2, height-1)*stride;
    cs->y = -3;
}

static void spatial_compose97i_dy_buffered(DWTContext *dsp, DWTCompose *cs, slice_buffer * sb, int width, int height, int stride_line){
    int y = cs->y;

    IDWTELEM *b0= cs->b0;
    IDWTELEM *b1= cs->b1;
    IDWTELEM *b2= cs->b2;
    IDWTELEM *b3= cs->b3;
    IDWTELEM *b4= slice_buffer_get_line(sb, mirror(y + 3, height - 1) * stride_line);
    IDWTELEM *b5= slice_buffer_get_line(sb, mirror(y + 4, height - 1) * stride_line);

    if(y>0 && y+4<height){
        dsp->vertical_compose97i(b0, b1, b2, b3, b4, b5, width);
    }else{
        if(y+3<(unsigned)height) vertical_compose97iL1(b3, b4, b5, width);
        if(y+2<(unsigned)height) vertical_compose97iH1(b2, b3, b4, width);
        if(y+1<(unsigned)height) vertical_compose97iL0(b1, b2, b3, width);
        if(y+0<(unsigned)height) vertical_compose97iH0(b0, b1, b2, width);
    }

    if(y-1<(unsigned)height) dsp->horizontal_compose97i(b0, width);
    if(y+0<(unsigned)height) dsp->horizontal_compose97i(b1, width);

    cs->b0=b2;
    cs->b1=b3;
    cs->b2=b4;
    cs->b3=b5;
    cs->y += 2;
}

static void spatial_compose97i_dy(DWTCompose *cs, IDWTELEM *buffer, int width, int height, int stride){
    int y = cs->y;
    IDWTELEM *b0= cs->b0;
    IDWTELEM *b1= cs->b1;
    IDWTELEM *b2= cs->b2;
    IDWTELEM *b3= cs->b3;
    IDWTELEM *b4= buffer + mirror(y+3, height-1)*stride;
    IDWTELEM *b5= buffer + mirror(y+4, height-1)*stride;

    if(y+3<(unsigned)height) vertical_compose97iL1(b3, b4, b5, width);
    if(y+2<(unsigned)height) vertical_compose97iH1(b2, b3, b4, width);
    if(y+1<(unsigned)height) vertical_compose97iL0(b1, b2, b3, width);
    if(y+0<(unsigned)height) vertical_compose97iH0(b0, b1, b2, width);

    if(y-1<(unsigned)height) ff_snow_horizontal_compose97i(b0, width);
    if(y+0<(unsigned)height) ff_snow_horizontal_compose97i(b1, width);

    cs->b0=b2;
    cs->b1=b3;
    cs->b2=b4;
    cs->b3=b5;
    cs->y += 2;
}

static void av_unused spatial_compose97i(IDWTELEM *buffer, int width, int height, int stride){
    DWTCompose cs;
    spatial_compose97i_init(&cs, buffer, height, stride);
    while(cs.y <= height)
        spatial_compose97i_dy(&cs, buffer, width, height, stride);
}

void ff_spatial_idwt_buffered_init(DWTCompose *cs, slice_buffer * sb, int width, int height, int stride_line, int type, int decomposition_count){
    int level;
    for(level=decomposition_count-1; level>=0; level--){
        switch(type){
        case DWT_97: spatial_compose97i_buffered_init(cs+level, sb, height>>level, stride_line<<level); break;
        case DWT_53: spatial_compose53i_buffered_init(cs+level, sb, height>>level, stride_line<<level); break;
        }
    }
}

void ff_spatial_idwt_buffered_slice(DWTContext *dsp, DWTCompose *cs, slice_buffer * slice_buf, int width, int height, int stride_line, int type, int decomposition_count, int y){
    const int support = type==1 ? 3 : 5;
    int level;
    if(type==2) return;

    for(level=decomposition_count-1; level>=0; level--){
        while(cs[level].y <= FFMIN((y>>level)+support, height>>level)){
            switch(type){
            case DWT_97: spatial_compose97i_dy_buffered(dsp, cs+level, slice_buf, width>>level, height>>level, stride_line<<level);
                break;
            case DWT_53: spatial_compose53i_dy_buffered(cs+level, slice_buf, width>>level, height>>level, stride_line<<level);
                break;
            }
        }
    }
}

703
static void ff_spatial_idwt_init(DWTCompose *cs, IDWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count){
704 705 706 707 708 709 710 711 712
    int level;
    for(level=decomposition_count-1; level>=0; level--){
        switch(type){
        case DWT_97: spatial_compose97i_init(cs+level, buffer, height>>level, stride<<level); break;
        case DWT_53: spatial_compose53i_init(cs+level, buffer, height>>level, stride<<level); break;
        }
    }
}

713
static void ff_spatial_idwt_slice(DWTCompose *cs, IDWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count, int y){
714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822
    const int support = type==1 ? 3 : 5;
    int level;
    if(type==2) return;

    for(level=decomposition_count-1; level>=0; level--){
        while(cs[level].y <= FFMIN((y>>level)+support, height>>level)){
            switch(type){
            case DWT_97: spatial_compose97i_dy(cs+level, buffer, width>>level, height>>level, stride<<level);
                break;
            case DWT_53: spatial_compose53i_dy(cs+level, buffer, width>>level, height>>level, stride<<level);
                break;
            }
        }
    }
}

void ff_spatial_idwt(IDWTELEM *buffer, int width, int height, int stride, int type, int decomposition_count){
        DWTCompose cs[MAX_DECOMPOSITIONS];
        int y;
        ff_spatial_idwt_init(cs, buffer, width, height, stride, type, decomposition_count);
        for(y=0; y<height; y+=4)
            ff_spatial_idwt_slice(cs, buffer, width, height, stride, type, decomposition_count, y);
}

static inline int w_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int w, int h, int type){
    int s, i, j;
    const int dec_count= w==8 ? 3 : 4;
    int tmp[32*32];
    int level, ori;
    static const int scale[2][2][4][4]={
      {
        {
            // 9/7 8x8 dec=3
            {268, 239, 239, 213},
            {  0, 224, 224, 152},
            {  0, 135, 135, 110},
        },{
            // 9/7 16x16 or 32x32 dec=4
            {344, 310, 310, 280},
            {  0, 320, 320, 228},
            {  0, 175, 175, 136},
            {  0, 129, 129, 102},
        }
      },{
        {
            // 5/3 8x8 dec=3
            {275, 245, 245, 218},
            {  0, 230, 230, 156},
            {  0, 138, 138, 113},
        },{
            // 5/3 16x16 or 32x32 dec=4
            {352, 317, 317, 286},
            {  0, 328, 328, 233},
            {  0, 180, 180, 140},
            {  0, 132, 132, 105},
        }
      }
    };

    for (i = 0; i < h; i++) {
        for (j = 0; j < w; j+=4) {
            tmp[32*i+j+0] = (pix1[j+0] - pix2[j+0])<<4;
            tmp[32*i+j+1] = (pix1[j+1] - pix2[j+1])<<4;
            tmp[32*i+j+2] = (pix1[j+2] - pix2[j+2])<<4;
            tmp[32*i+j+3] = (pix1[j+3] - pix2[j+3])<<4;
        }
        pix1 += line_size;
        pix2 += line_size;
    }

    ff_spatial_dwt(tmp, w, h, 32, type, dec_count);

    s=0;
    assert(w==h);
    for(level=0; level<dec_count; level++){
        for(ori= level ? 1 : 0; ori<4; ori++){
            int size= w>>(dec_count-level);
            int sx= (ori&1) ? size : 0;
            int stride= 32<<(dec_count-level);
            int sy= (ori&2) ? stride>>1 : 0;

            for(i=0; i<size; i++){
                for(j=0; j<size; j++){
                    int v= tmp[sx + sy + i*stride + j] * scale[type][dec_count-3][level][ori];
                    s += FFABS(v);
                }
            }
        }
    }
    assert(s>=0);
    return s>>9;
}

static int w53_8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
    return w_c(v, pix1, pix2, line_size,  8, h, 1);
}

static int w97_8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
    return w_c(v, pix1, pix2, line_size,  8, h, 0);
}

static int w53_16_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
    return w_c(v, pix1, pix2, line_size, 16, h, 1);
}

static int w97_16_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
    return w_c(v, pix1, pix2, line_size, 16, h, 0);
}

823
int ff_w53_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
824 825 826
    return w_c(v, pix1, pix2, line_size, 32, h, 1);
}

827
int ff_w97_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){
828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844
    return w_c(v, pix1, pix2, line_size, 32, h, 0);
}

void ff_dsputil_init_dwt(DSPContext *c)
{
    c->w53[0]= w53_16_c;
    c->w53[1]= w53_8_c;
    c->w97[0]= w97_16_c;
    c->w97[1]= w97_8_c;
}

void ff_dwt_init(DWTContext *c)
{
    c->vertical_compose97i = ff_snow_vertical_compose97i;
    c->horizontal_compose97i = ff_snow_horizontal_compose97i;
    c->inner_add_yblock = ff_snow_inner_add_yblock;

845
    if (HAVE_MMX) ff_dwt_init_x86(c);
846
}
847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351


static av_always_inline
void interleave(IDWTELEM *dst, IDWTELEM *src0, IDWTELEM *src1, int w2, int add, int shift)
{
    int i;
    for (i = 0; i < w2; i++) {
        dst[2*i  ] = (src0[i] + add) >> shift;
        dst[2*i+1] = (src1[i] + add) >> shift;
    }
}

static void horizontal_compose_dirac53i(IDWTELEM *b, IDWTELEM *temp, int w)
{
    const int w2 = w >> 1;
    int x;

    temp[0] = COMPOSE_53iL0(b[w2], b[0], b[w2]);
    for (x = 1; x < w2; x++) {
        temp[x     ] = COMPOSE_53iL0     (b[x+w2-1], b[x     ], b[x+w2]);
        temp[x+w2-1] = COMPOSE_DIRAC53iH0(temp[x-1], b[x+w2-1], temp[x]);
    }
    temp[w-1] = COMPOSE_DIRAC53iH0(temp[w2-1], b[w-1], temp[w2-1]);

    interleave(b, temp, temp+w2, w2, 1, 1);
}

static void horizontal_compose_dd97i(IDWTELEM *b, IDWTELEM *tmp, int w)
{
    const int w2 = w >> 1;
    int x;

    tmp[0] = COMPOSE_53iL0(b[w2], b[0], b[w2]);
    for (x = 1; x < w2; x++)
        tmp[x] = COMPOSE_53iL0(b[x+w2-1], b[x], b[x+w2]);

    // extend the edges
    tmp[-1]   = tmp[0];
    tmp[w2+1] = tmp[w2] = tmp[w2-1];

    for (x = 0; x < w2; x++) {
        b[2*x  ] = (tmp[x] + 1)>>1;
        b[2*x+1] = (COMPOSE_DD97iH0(tmp[x-1], tmp[x], b[x+w2], tmp[x+1], tmp[x+2]) + 1)>>1;
    }
}

static void horizontal_compose_dd137i(IDWTELEM *b, IDWTELEM *tmp, int w)
{
    const int w2 = w >> 1;
    int x;

    tmp[0] = COMPOSE_DD137iL0(b[w2], b[w2], b[0], b[w2  ], b[w2+1]);
    tmp[1] = COMPOSE_DD137iL0(b[w2], b[w2], b[1], b[w2+1], b[w2+2]);
    for (x = 2; x < w2-1; x++)
        tmp[x] = COMPOSE_DD137iL0(b[x+w2-2], b[x+w2-1], b[x], b[x+w2], b[x+w2+1]);
    tmp[w2-1] = COMPOSE_DD137iL0(b[w-3], b[w-2], b[w2-1], b[w-1], b[w-1]);

    // extend the edges
    tmp[-1]   = tmp[0];
    tmp[w2+1] = tmp[w2] = tmp[w2-1];

    for (x = 0; x < w2; x++) {
        b[2*x  ] = (tmp[x] + 1)>>1;
        b[2*x+1] = (COMPOSE_DD97iH0(tmp[x-1], tmp[x], b[x+w2], tmp[x+1], tmp[x+2]) + 1)>>1;
    }
}

static av_always_inline
void horizontal_compose_haari(IDWTELEM *b, IDWTELEM *temp, int w, int shift)
{
    const int w2 = w >> 1;
    int x;

    for (x = 0; x < w2; x++) {
        temp[x   ] = COMPOSE_HAARiL0(b[x   ], b[x+w2]);
        temp[x+w2] = COMPOSE_HAARiH0(b[x+w2], temp[x]);
    }

    interleave(b, temp, temp+w2, w2, shift, shift);
}

static void horizontal_compose_haar0i(IDWTELEM *b, IDWTELEM *temp, int w)
{
    horizontal_compose_haari(b, temp, w, 0);
}

static void horizontal_compose_haar1i(IDWTELEM *b, IDWTELEM *temp, int w)
{
    horizontal_compose_haari(b, temp, w, 1);
}

static void horizontal_compose_fidelityi(IDWTELEM *b, IDWTELEM *tmp, int w)
{
    const int w2 = w >> 1;
    int i, x;
    IDWTELEM v[8];

    for (x = 0; x < w2; x++) {
        for (i = 0; i < 8; i++)
            v[i] = b[av_clip(x-3+i, 0, w2-1)];
        tmp[x] = COMPOSE_FIDELITYiH0(v[0], v[1], v[2], v[3], b[x+w2], v[4], v[5], v[6], v[7]);
    }

    for (x = 0; x < w2; x++) {
        for (i = 0; i < 8; i++)
            v[i] = tmp[av_clip(x-4+i, 0, w2-1)];
        tmp[x+w2] = COMPOSE_FIDELITYiL0(v[0], v[1], v[2], v[3], b[x], v[4], v[5], v[6], v[7]);
    }

    interleave(b, tmp+w2, tmp, w2, 0, 0);
}

static void horizontal_compose_daub97i(IDWTELEM *b, IDWTELEM *temp, int w)
{
    const int w2 = w >> 1;
    int x, b0, b1, b2;

    temp[0] = COMPOSE_DAUB97iL1(b[w2], b[0], b[w2]);
    for (x = 1; x < w2; x++) {
        temp[x     ] = COMPOSE_DAUB97iL1(b[x+w2-1], b[x     ], b[x+w2]);
        temp[x+w2-1] = COMPOSE_DAUB97iH1(temp[x-1], b[x+w2-1], temp[x]);
    }
    temp[w-1] = COMPOSE_DAUB97iH1(temp[w2-1], b[w-1], temp[w2-1]);

    // second stage combined with interleave and shift
    b0 = b2 = COMPOSE_DAUB97iL0(temp[w2], temp[0], temp[w2]);
    b[0] = (b0 + 1) >> 1;
    for (x = 1; x < w2; x++) {
        b2 = COMPOSE_DAUB97iL0(temp[x+w2-1], temp[x     ], temp[x+w2]);
        b1 = COMPOSE_DAUB97iH0(          b0, temp[x+w2-1], b2        );
        b[2*x-1] = (b1 + 1) >> 1;
        b[2*x  ] = (b2 + 1) >> 1;
        b0 = b2;
    }
    b[w-1] = (COMPOSE_DAUB97iH0(b2, temp[w-1], b2) + 1) >> 1;
}

static void vertical_compose_dirac53iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
{
    int i;

    for(i=0; i<width; i++){
        b1[i] = COMPOSE_DIRAC53iH0(b0[i], b1[i], b2[i]);
    }
}

static void vertical_compose_dd97iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
                                  IDWTELEM *b3, IDWTELEM *b4, int width)
{
    int i;

    for(i=0; i<width; i++){
        b2[i] = COMPOSE_DD97iH0(b0[i], b1[i], b2[i], b3[i], b4[i]);
    }
}

static void vertical_compose_dd137iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2,
                                      IDWTELEM *b3, IDWTELEM *b4, int width)
{
    int i;

    for(i=0; i<width; i++){
        b2[i] = COMPOSE_DD137iL0(b0[i], b1[i], b2[i], b3[i], b4[i]);
    }
}

static void vertical_compose_haar(IDWTELEM *b0, IDWTELEM *b1, int width)
{
    int i;

    for (i = 0; i < width; i++) {
        b0[i] = COMPOSE_HAARiL0(b0[i], b1[i]);
        b1[i] = COMPOSE_HAARiH0(b1[i], b0[i]);
    }
}

static void vertical_compose_fidelityiH0(IDWTELEM *dst, IDWTELEM *b[8], int width)
{
    int i;

    for(i=0; i<width; i++){
        dst[i] = COMPOSE_FIDELITYiH0(b[0][i], b[1][i], b[2][i], b[3][i], dst[i], b[4][i], b[5][i], b[6][i], b[7][i]);
    }
}

static void vertical_compose_fidelityiL0(IDWTELEM *dst, IDWTELEM *b[8], int width)
{
    int i;

    for(i=0; i<width; i++){
        dst[i] = COMPOSE_FIDELITYiL0(b[0][i], b[1][i], b[2][i], b[3][i], dst[i], b[4][i], b[5][i], b[6][i], b[7][i]);
    }
}

static void vertical_compose_daub97iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
{
    int i;

    for(i=0; i<width; i++){
        b1[i] = COMPOSE_DAUB97iH0(b0[i], b1[i], b2[i]);
    }
}

static void vertical_compose_daub97iH1(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
{
    int i;

    for(i=0; i<width; i++){
        b1[i] = COMPOSE_DAUB97iH1(b0[i], b1[i], b2[i]);
    }
}

static void vertical_compose_daub97iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
{
    int i;

    for(i=0; i<width; i++){
        b1[i] = COMPOSE_DAUB97iL0(b0[i], b1[i], b2[i]);
    }
}

static void vertical_compose_daub97iL1(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width)
{
    int i;

    for(i=0; i<width; i++){
        b1[i] = COMPOSE_DAUB97iL1(b0[i], b1[i], b2[i]);
    }
}


static void spatial_compose_dd97i_dy(DWTContext *d, int level, int width, int height, int stride)
{
    vertical_compose_3tap vertical_compose_l0 = d->vertical_compose_l0;
    vertical_compose_5tap vertical_compose_h0 = d->vertical_compose_h0;
    DWTCompose *cs = d->cs + level;

    int i, y = cs->y;
    IDWTELEM *b[8];
    for (i = 0; i < 6; i++)
        b[i] = cs->b[i];
    b[6] = d->buffer + av_clip(y+5, 0, height-2)*stride;
    b[7] = d->buffer + av_clip(y+6, 1, height-1)*stride;

        if(y+5<(unsigned)height) vertical_compose_l0(      b[5], b[6], b[7],       width);
        if(y+1<(unsigned)height) vertical_compose_h0(b[0], b[2], b[3], b[4], b[6], width);

        if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
        if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);

    for (i = 0; i < 6; i++)
        cs->b[i] = b[i+2];
    cs->y += 2;
}

static void spatial_compose_dirac53i_dy(DWTContext *d, int level, int width, int height, int stride)
{
    vertical_compose_3tap vertical_compose_l0 = d->vertical_compose_l0;
    vertical_compose_3tap vertical_compose_h0 = d->vertical_compose_h0;
    DWTCompose *cs = d->cs + level;

    int y= cs->y;
    IDWTELEM *b[4] = { cs->b[0], cs->b[1] };
    b[2] = d->buffer + mirror(y+1, height-1)*stride;
    b[3] = d->buffer + mirror(y+2, height-1)*stride;

        if(y+1<(unsigned)height) vertical_compose_l0(b[1], b[2], b[3], width);
        if(y+0<(unsigned)height) vertical_compose_h0(b[0], b[1], b[2], width);

        if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
        if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);

    cs->b[0] = b[2];
    cs->b[1] = b[3];
    cs->y += 2;
}


static void spatial_compose_dd137i_dy(DWTContext *d, int level, int width, int height, int stride)
{
    vertical_compose_5tap vertical_compose_l0 = d->vertical_compose_l0;
    vertical_compose_5tap vertical_compose_h0 = d->vertical_compose_h0;
    DWTCompose *cs = d->cs + level;

    int i, y = cs->y;
    IDWTELEM *b[10];
    for (i = 0; i < 8; i++)
        b[i] = cs->b[i];
    b[8] = d->buffer + av_clip(y+7, 0, height-2)*stride;
    b[9] = d->buffer + av_clip(y+8, 1, height-1)*stride;

        if(y+5<(unsigned)height) vertical_compose_l0(b[3], b[5], b[6], b[7], b[9], width);
        if(y+1<(unsigned)height) vertical_compose_h0(b[0], b[2], b[3], b[4], b[6], width);

        if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
        if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);

    for (i = 0; i < 8; i++)
        cs->b[i] = b[i+2];
    cs->y += 2;
}

// haar makes the assumption that height is even (always true for dirac)
static void spatial_compose_haari_dy(DWTContext *d, int level, int width, int height, int stride)
{
    vertical_compose_2tap vertical_compose = d->vertical_compose;
    int y = d->cs[level].y;
    IDWTELEM *b0 = d->buffer + (y-1)*stride;
    IDWTELEM *b1 = d->buffer + (y  )*stride;

    vertical_compose(b0, b1, width);
    d->horizontal_compose(b0, d->temp, width);
    d->horizontal_compose(b1, d->temp, width);

    d->cs[level].y += 2;
}

// Don't do sliced idwt for fidelity; the 9 tap filter makes it a bit annoying
// Fortunately, this filter isn't used in practice.
static void spatial_compose_fidelity(DWTContext *d, int level, int width, int height, int stride)
{
    vertical_compose_9tap vertical_compose_l0 = d->vertical_compose_l0;
    vertical_compose_9tap vertical_compose_h0 = d->vertical_compose_h0;
    int i, y;
    IDWTELEM *b[8];

    for (y = 1; y < height; y += 2) {
        for (i = 0; i < 8; i++)
            b[i] = d->buffer + av_clip((y-7 + 2*i), 0, height-2)*stride;
        vertical_compose_h0(d->buffer + y*stride, b, width);
    }

    for (y = 0; y < height; y += 2) {
        for (i = 0; i < 8; i++)
            b[i] = d->buffer + av_clip((y-7 + 2*i), 1, height-1)*stride;
        vertical_compose_l0(d->buffer + y*stride, b, width);
    }

    for (y = 0; y < height; y++)
        d->horizontal_compose(d->buffer + y*stride, d->temp, width);

    d->cs[level].y = height+1;
}

static void spatial_compose_daub97i_dy(DWTContext *d, int level, int width, int height, int stride)
{
    vertical_compose_3tap vertical_compose_l0 = d->vertical_compose_l0;
    vertical_compose_3tap vertical_compose_h0 = d->vertical_compose_h0;
    vertical_compose_3tap vertical_compose_l1 = d->vertical_compose_l1;
    vertical_compose_3tap vertical_compose_h1 = d->vertical_compose_h1;
    DWTCompose *cs = d->cs + level;

    int i, y = cs->y;
    IDWTELEM *b[6];
    for (i = 0; i < 4; i++)
        b[i] = cs->b[i];
    b[4] = d->buffer + mirror(y+3, height-1)*stride;
    b[5] = d->buffer + mirror(y+4, height-1)*stride;

        if(y+3<(unsigned)height) vertical_compose_l1(b[3], b[4], b[5], width);
        if(y+2<(unsigned)height) vertical_compose_h1(b[2], b[3], b[4], width);
        if(y+1<(unsigned)height) vertical_compose_l0(b[1], b[2], b[3], width);
        if(y+0<(unsigned)height) vertical_compose_h0(b[0], b[1], b[2], width);

        if(y-1<(unsigned)height) d->horizontal_compose(b[0], d->temp, width);
        if(y+0<(unsigned)height) d->horizontal_compose(b[1], d->temp, width);

    for (i = 0; i < 4; i++)
        cs->b[i] = b[i+2];
    cs->y += 2;
}


static void spatial_compose97i_init2(DWTCompose *cs, IDWTELEM *buffer, int height, int stride)
{
    cs->b[0] = buffer + mirror(-3-1, height-1)*stride;
    cs->b[1] = buffer + mirror(-3  , height-1)*stride;
    cs->b[2] = buffer + mirror(-3+1, height-1)*stride;
    cs->b[3] = buffer + mirror(-3+2, height-1)*stride;
    cs->y = -3;
}

static void spatial_compose53i_init2(DWTCompose *cs, IDWTELEM *buffer, int height, int stride)
{
    cs->b[0] = buffer + mirror(-1-1, height-1)*stride;
    cs->b[1] = buffer + mirror(-1  , height-1)*stride;
    cs->y = -1;
}

static void spatial_compose_dd97i_init(DWTCompose *cs, IDWTELEM *buffer, int height, int stride)
{
    cs->b[0] = buffer + av_clip(-5-1, 0, height-2)*stride;
    cs->b[1] = buffer + av_clip(-5  , 1, height-1)*stride;
    cs->b[2] = buffer + av_clip(-5+1, 0, height-2)*stride;
    cs->b[3] = buffer + av_clip(-5+2, 1, height-1)*stride;
    cs->b[4] = buffer + av_clip(-5+3, 0, height-2)*stride;
    cs->b[5] = buffer + av_clip(-5+4, 1, height-1)*stride;
    cs->y = -5;
}

static void spatial_compose_dd137i_init(DWTCompose *cs, IDWTELEM *buffer, int height, int stride)
{
    cs->b[0] = buffer + av_clip(-5-1, 0, height-2)*stride;
    cs->b[1] = buffer + av_clip(-5  , 1, height-1)*stride;
    cs->b[2] = buffer + av_clip(-5+1, 0, height-2)*stride;
    cs->b[3] = buffer + av_clip(-5+2, 1, height-1)*stride;
    cs->b[4] = buffer + av_clip(-5+3, 0, height-2)*stride;
    cs->b[5] = buffer + av_clip(-5+4, 1, height-1)*stride;
    cs->b[6] = buffer + av_clip(-5+5, 0, height-2)*stride;
    cs->b[7] = buffer + av_clip(-5+6, 1, height-1)*stride;
    cs->y = -5;
}

int ff_spatial_idwt_init2(DWTContext *d, IDWTELEM *buffer, int width, int height,
                          int stride, enum dwt_type type, int decomposition_count,
                          IDWTELEM *temp)
{
    int level;

    d->buffer = buffer;
    d->width = width;
    d->height = height;
    d->stride = stride;
    d->decomposition_count = decomposition_count;
    d->temp = temp + 8;

    for(level=decomposition_count-1; level>=0; level--){
        int hl = height >> level;
        int stride_l = stride << level;

        switch(type){
        case DWT_DIRAC_DD9_7:
            spatial_compose_dd97i_init(d->cs+level, buffer, hl, stride_l);
            break;
        case DWT_DIRAC_LEGALL5_3:
            spatial_compose53i_init2(d->cs+level, buffer, hl, stride_l);
            break;
        case DWT_DIRAC_DD13_7:
            spatial_compose_dd137i_init(d->cs+level, buffer, hl, stride_l);
            break;
        case DWT_DIRAC_HAAR0:
        case DWT_DIRAC_HAAR1:
            d->cs[level].y = 1;
            break;
        case DWT_DIRAC_DAUB9_7:
            spatial_compose97i_init2(d->cs+level, buffer, hl, stride_l);
            break;
        default:
            d->cs[level].y = 0;
            break;
        }
    }

    switch (type) {
    case DWT_DIRAC_DD9_7:
        d->spatial_compose = spatial_compose_dd97i_dy;
        d->vertical_compose_l0 = vertical_compose53iL0;
        d->vertical_compose_h0 = vertical_compose_dd97iH0;
        d->horizontal_compose = horizontal_compose_dd97i;
        d->support = 7;
        break;
    case DWT_DIRAC_LEGALL5_3:
        d->spatial_compose = spatial_compose_dirac53i_dy;
        d->vertical_compose_l0 = vertical_compose53iL0;
        d->vertical_compose_h0 = vertical_compose_dirac53iH0;
        d->horizontal_compose = horizontal_compose_dirac53i;
        d->support = 3;
        break;
    case DWT_DIRAC_DD13_7:
        d->spatial_compose = spatial_compose_dd137i_dy;
        d->vertical_compose_l0 = vertical_compose_dd137iL0;
        d->vertical_compose_h0 = vertical_compose_dd97iH0;
        d->horizontal_compose = horizontal_compose_dd137i;
        d->support = 7;
        break;
    case DWT_DIRAC_HAAR0:
    case DWT_DIRAC_HAAR1:
        d->spatial_compose = spatial_compose_haari_dy;
        d->vertical_compose = vertical_compose_haar;
        if (type == DWT_DIRAC_HAAR0)
            d->horizontal_compose = horizontal_compose_haar0i;
        else
            d->horizontal_compose = horizontal_compose_haar1i;
        d->support = 1;
        break;
    case DWT_DIRAC_FIDELITY:
        d->spatial_compose = spatial_compose_fidelity;
        d->vertical_compose_l0 = vertical_compose_fidelityiL0;
        d->vertical_compose_h0 = vertical_compose_fidelityiH0;
        d->horizontal_compose = horizontal_compose_fidelityi;
        break;
    case DWT_DIRAC_DAUB9_7:
        d->spatial_compose = spatial_compose_daub97i_dy;
        d->vertical_compose_l0 = vertical_compose_daub97iL0;
        d->vertical_compose_h0 = vertical_compose_daub97iH0;
        d->vertical_compose_l1 = vertical_compose_daub97iL1;
        d->vertical_compose_h1 = vertical_compose_daub97iH1;
        d->horizontal_compose = horizontal_compose_daub97i;
        d->support = 5;
        break;
    default:
        av_log(NULL, AV_LOG_ERROR, "Unknown wavelet type %d\n", type);
        return -1;
    }

1352
    if (HAVE_MMX) ff_spatial_idwt_init_mmx(d, type);
1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384

    return 0;
}

void ff_spatial_idwt_slice2(DWTContext *d, int y)
{
    int level, support = d->support;

    for (level = d->decomposition_count-1; level >= 0; level--) {
        int wl = d->width  >> level;
        int hl = d->height >> level;
        int stride_l = d->stride << level;

        while (d->cs[level].y <= FFMIN((y>>level)+support, hl))
            d->spatial_compose(d, level, wl, hl, stride_l);
    }
}

int ff_spatial_idwt2(IDWTELEM *buffer, int width, int height, int stride,
                     enum dwt_type type, int decomposition_count, IDWTELEM *temp)
{
    DWTContext d;
    int y;

    if (ff_spatial_idwt_init2(&d, buffer, width, height, stride, type, decomposition_count, temp))
        return -1;

    for (y = 0; y < d.height; y += 4)
        ff_spatial_idwt_slice2(&d, y);

    return 0;
}