mpegvideo.c 44.1 KB
Newer Older
F
Fabrice Bellard 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26
/*
 * The simplest mpeg encoder (well, it was the simplest!)
 * Copyright (c) 2000,2001 Gerard Lantau.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program 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 General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */
#include <stdlib.h>
#include <stdio.h>
#include <math.h>
#include <string.h>
#include "avcodec.h"
#include "dsputil.h"
#include "mpegvideo.h"

27 28 29 30
#ifdef USE_FASTMEMCPY
#include "fastmemcpy.h"
#endif

31 32 33 34 35 36 37 38 39 40 41
static void encode_picture(MpegEncContext *s, int picture_number);
static void rate_control_init(MpegEncContext *s);
static int rate_estimate_qscale(MpegEncContext *s);
static void dct_unquantize_mpeg1_c(MpegEncContext *s, 
                                   DCTELEM *block, int n, int qscale);
static void dct_unquantize_h263_c(MpegEncContext *s, 
                                  DCTELEM *block, int n, int qscale);
static int dct_quantize(MpegEncContext *s, DCTELEM *block, int n, int qscale);
static int dct_quantize_mmx(MpegEncContext *s, 
                            DCTELEM *block, int n,
                            int qscale);
F
Fabrice Bellard 已提交
42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83
#define EDGE_WIDTH 16

/* enable all paranoid tests for rounding, overflows, etc... */
//#define PARANOID

//#define DEBUG

/* for jpeg fast DCT */
#define CONST_BITS 14

static const unsigned short aanscales[64] = {
    /* precomputed values scaled up by 14 bits */
    16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
    22725, 31521, 29692, 26722, 22725, 17855, 12299,  6270,
    21407, 29692, 27969, 25172, 21407, 16819, 11585,  5906,
    19266, 26722, 25172, 22654, 19266, 15137, 10426,  5315,
    16384, 22725, 21407, 19266, 16384, 12873,  8867,  4520,
    12873, 17855, 16819, 15137, 12873, 10114,  6967,  3552,
    8867, 12299, 11585, 10426,  8867,  6967,  4799,  2446,
    4520,  6270,  5906,  5315,  4520,  3552,  2446,  1247
};

static UINT8 h263_chroma_roundtab[16] = {
    0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2,
};

/* default motion estimation */
int motion_estimation_method = ME_LOG;

/* XXX: should use variable shift ? */
#define QMAT_SHIFT_MMX 19
#define QMAT_SHIFT 25

static void convert_matrix(int *qmat, const UINT16 *quant_matrix, int qscale)
{
    int i;

    if (av_fdct == jpeg_fdct_ifast) {
        for(i=0;i<64;i++) {
            /* 16 <= qscale * quant_matrix[i] <= 7905 */
            /* 19952 <= aanscales[i] * qscale * quant_matrix[i] <= 249205026 */
            
F
Fabrice Bellard 已提交
84 85
            qmat[i] = (int)((UINT64_C(1) << (QMAT_SHIFT + 11)) / 
                            (aanscales[i] * qscale * quant_matrix[i]));
F
Fabrice Bellard 已提交
86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103
        }
    } else {
        for(i=0;i<64;i++) {
            /* We can safely suppose that 16 <= quant_matrix[i] <= 255
               So 16 <= qscale * quant_matrix[i] <= 7905
               so (1 << QMAT_SHIFT) / 16 >= qmat[i] >= (1 << QMAT_SHIFT) / 7905
            */
            qmat[i] = (1 << QMAT_SHIFT_MMX) / (qscale * quant_matrix[i]);
        }
    }
}

/* init common structure for both encoder and decoder */
int MPV_common_init(MpegEncContext *s)
{
    int c_size, i;
    UINT8 *pict;

104 105 106 107 108 109 110
    if (s->out_format == FMT_H263) 
        s->dct_unquantize = dct_unquantize_h263_c;
    else
        s->dct_unquantize = dct_unquantize_mpeg1_c;
        
#ifdef HAVE_MMX
    MPV_common_init_mmx(s);
N
Nick Kurshev 已提交
111
#endif
F
Fabrice Bellard 已提交
112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182
    s->mb_width = (s->width + 15) / 16;
    s->mb_height = (s->height + 15) / 16;
    s->linesize = s->mb_width * 16 + 2 * EDGE_WIDTH;

    for(i=0;i<3;i++) {
        int w, h, shift, pict_start;

        w = s->linesize;
        h = s->mb_height * 16 + 2 * EDGE_WIDTH;
        shift = (i == 0) ? 0 : 1;
        c_size = (w >> shift) * (h >> shift);
        pict_start = (w >> shift) * (EDGE_WIDTH >> shift) + (EDGE_WIDTH >> shift);

        pict = av_mallocz(c_size);
        if (pict == NULL)
            goto fail;
        s->last_picture_base[i] = pict;
        s->last_picture[i] = pict + pict_start;
    
        pict = av_mallocz(c_size);
        if (pict == NULL)
            goto fail;
        s->next_picture_base[i] = pict;
        s->next_picture[i] = pict + pict_start;

        if (s->has_b_frames) {
            pict = av_mallocz(c_size);
            if (pict == NULL) 
                goto fail;
            s->aux_picture_base[i] = pict;
            s->aux_picture[i] = pict + pict_start;
        }
    }

    if (s->out_format == FMT_H263) {
        int size;
        /* MV prediction */
        size = (2 * s->mb_width + 2) * (2 * s->mb_height + 2);
        s->motion_val = malloc(size * 2 * sizeof(INT16));
        if (s->motion_val == NULL)
            goto fail;
        memset(s->motion_val, 0, size * 2 * sizeof(INT16));
    }

    if (s->h263_pred) {
        int y_size, c_size, i, size;
        
        /* dc values */

        y_size = (2 * s->mb_width + 2) * (2 * s->mb_height + 2);
        c_size = (s->mb_width + 2) * (s->mb_height + 2);
        size = y_size + 2 * c_size;
        s->dc_val[0] = malloc(size * sizeof(INT16));
        if (s->dc_val[0] == NULL)
            goto fail;
        s->dc_val[1] = s->dc_val[0] + y_size;
        s->dc_val[2] = s->dc_val[1] + c_size;
        for(i=0;i<size;i++)
            s->dc_val[0][i] = 1024;

        /* ac values */
        s->ac_val[0] = av_mallocz(size * sizeof(INT16) * 16);
        if (s->ac_val[0] == NULL)
            goto fail;
        s->ac_val[1] = s->ac_val[0] + y_size;
        s->ac_val[2] = s->ac_val[1] + c_size;
        
        /* cbp values */
        s->coded_block = av_mallocz(y_size);
        if (!s->coded_block)
            goto fail;
183
    }
184 185

        /* which mb is a intra block */
186
        s->mbintra_table = av_mallocz(s->mb_width * s->mb_height);
187 188
        if (!s->mbintra_table)
            goto fail;
189
        memset(s->mbintra_table, 1, s->mb_width * s->mb_height);
F
Fabrice Bellard 已提交
190 191 192
    /* default structure is frame */
    s->picture_structure = PICT_FRAME;

193 194 195 196 197 198 199
    /* init macroblock skip table */
    if (!s->encoding) {
        s->mbskip_table = av_mallocz(s->mb_width * s->mb_height);
        if (!s->mbskip_table)
            goto fail;
    }

F
Fabrice Bellard 已提交
200 201 202 203 204 205 206 207 208 209 210
    s->context_initialized = 1;
    return 0;
 fail:
    if (s->motion_val)
        free(s->motion_val);
    if (s->dc_val[0])
        free(s->dc_val[0]);
    if (s->ac_val[0])
        free(s->ac_val[0]);
    if (s->coded_block)
        free(s->coded_block);
211
    if (s->mbintra_table)
212
        { free(s->mbintra_table);s->mbintra_table=NULL; }
213 214
    if (s->mbskip_table)
        free(s->mbskip_table);
F
Fabrice Bellard 已提交
215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236
    for(i=0;i<3;i++) {
        if (s->last_picture_base[i])
            free(s->last_picture_base[i]);
        if (s->next_picture_base[i])
            free(s->next_picture_base[i]);
        if (s->aux_picture_base[i])
            free(s->aux_picture_base[i]);
    }
    return -1;
}

/* init common structure for both encoder and decoder */
void MPV_common_end(MpegEncContext *s)
{
    int i;

    if (s->motion_val)
        free(s->motion_val);
    if (s->h263_pred) {
        free(s->dc_val[0]);
        free(s->ac_val[0]);
        free(s->coded_block);
237
        { free(s->mbintra_table);s->mbintra_table=NULL; }
F
Fabrice Bellard 已提交
238
    }
239 240
    if (s->mbskip_table)
        free(s->mbskip_table);
F
Fabrice Bellard 已提交
241 242 243 244 245 246 247 248 249 250 251 252 253
    for(i=0;i<3;i++) {
        free(s->last_picture_base[i]);
        free(s->next_picture_base[i]);
        if (s->has_b_frames)
            free(s->aux_picture_base[i]);
    }
    s->context_initialized = 0;
}

/* init video encoder */
int MPV_encode_init(AVCodecContext *avctx)
{
    MpegEncContext *s = avctx->priv_data;
254
    int i;
F
Fabrice Bellard 已提交
255 256 257 258 259 260

    s->bit_rate = avctx->bit_rate;
    s->frame_rate = avctx->frame_rate;
    s->width = avctx->width;
    s->height = avctx->height;
    s->gop_size = avctx->gop_size;
261 262
    s->rtp_mode = avctx->rtp_mode;
    s->rtp_payload_size = avctx->rtp_payload_size;
263
    s->avctx = avctx;
264
    
F
Fabrice Bellard 已提交
265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285
    if (s->gop_size <= 1) {
        s->intra_only = 1;
        s->gop_size = 12;
    } else {
        s->intra_only = 0;
    }
    s->full_search = motion_estimation_method;

    s->fixed_qscale = (avctx->flags & CODEC_FLAG_QSCALE);

    switch(avctx->codec->id) {
    case CODEC_ID_MPEG1VIDEO:
        s->out_format = FMT_MPEG1;
        break;
    case CODEC_ID_MJPEG:
        s->out_format = FMT_MJPEG;
        s->intra_only = 1; /* force intra only for jpeg */
        if (mjpeg_init(s) < 0)
            return -1;
        break;
    case CODEC_ID_H263:
A
Arpi 已提交
286 287
        if (h263_get_picture_format(s->width, s->height) == 7){
	    printf("Input picture size isn't suitable for h263 codec! try h263+\n");
F
Fabrice Bellard 已提交
288
            return -1;
A
Arpi 已提交
289
	}
F
Fabrice Bellard 已提交
290 291 292 293
        s->out_format = FMT_H263;
        break;
    case CODEC_ID_H263P:
        s->out_format = FMT_H263;
294 295
        s->rtp_mode = 1;
        s->rtp_payload_size = 1200; 
F
Fabrice Bellard 已提交
296
        s->h263_plus = 1;
297
        s->unrestricted_mv = 1;
298 299 300 301
        
        /* These are just to be sure */
        s->umvplus = 0;
        s->umvplus_dec = 0;
F
Fabrice Bellard 已提交
302 303 304 305 306
        break;
    case CODEC_ID_RV10:
        s->out_format = FMT_H263;
        s->h263_rv10 = 1;
        break;
F
Fabrice Bellard 已提交
307
    case CODEC_ID_MPEG4:
F
Fabrice Bellard 已提交
308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324
        s->out_format = FMT_H263;
        s->h263_pred = 1;
        s->unrestricted_mv = 1;
        break;
    case CODEC_ID_MSMPEG4:
        s->out_format = FMT_H263;
        s->h263_msmpeg4 = 1;
        s->h263_pred = 1;
        s->unrestricted_mv = 1;
        break;
    default:
        return -1;
    }

    if (s->out_format == FMT_H263)
        h263_encode_init_vlc(s);

325 326
    s->encoding = 1;

F
Fabrice Bellard 已提交
327 328 329 330
    /* init */
    if (MPV_common_init(s) < 0)
        return -1;
    
331 332 333 334 335 336
    /* init default q matrix */
    for(i=0;i<64;i++) {
        s->intra_matrix[i] = default_intra_matrix[i];
        s->non_intra_matrix[i] = default_non_intra_matrix[i];
    }

F
Fabrice Bellard 已提交
337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394
    /* rate control init */
    rate_control_init(s);

    s->picture_number = 0;
    s->fake_picture_number = 0;
    /* motion detector init */
    s->f_code = 1;

    return 0;
}

int MPV_encode_end(AVCodecContext *avctx)
{
    MpegEncContext *s = avctx->priv_data;

#ifdef STATS
    print_stats();
#endif
    MPV_common_end(s);
    if (s->out_format == FMT_MJPEG)
        mjpeg_close(s);
    return 0;
}

/* draw the edges of width 'w' of an image of size width, height */
static void draw_edges(UINT8 *buf, int wrap, int width, int height, int w)
{
    UINT8 *ptr, *last_line;
    int i;

    last_line = buf + (height - 1) * wrap;
    for(i=0;i<w;i++) {
        /* top and bottom */
        memcpy(buf - (i + 1) * wrap, buf, width);
        memcpy(last_line + (i + 1) * wrap, last_line, width);
    }
    /* left and right */
    ptr = buf;
    for(i=0;i<height;i++) {
        memset(ptr - w, ptr[0], w);
        memset(ptr + width, ptr[width-1], w);
        ptr += wrap;
    }
    /* corners */
    for(i=0;i<w;i++) {
        memset(buf - (i + 1) * wrap - w, buf[0], w); /* top left */
        memset(buf - (i + 1) * wrap + width, buf[width-1], w); /* top right */
        memset(last_line + (i + 1) * wrap - w, last_line[0], w); /* top left */
        memset(last_line + (i + 1) * wrap + width, last_line[width-1], w); /* top right */
    }
}

/* generic function for encode/decode called before a frame is coded/decoded */
void MPV_frame_start(MpegEncContext *s)
{
    int i;
    UINT8 *tmp;

395
    s->mb_skiped = 0;
F
Fabrice Bellard 已提交
396 397 398 399 400 401 402 403 404 405 406 407 408 409
    if (s->pict_type == B_TYPE) {
        for(i=0;i<3;i++) {
            s->current_picture[i] = s->aux_picture[i];
        }
    } else {
        for(i=0;i<3;i++) {
            /* swap next and last */
            tmp = s->last_picture[i];
            s->last_picture[i] = s->next_picture[i];
            s->next_picture[i] = tmp;
            s->current_picture[i] = tmp;
        }
    }
}
410

F
Fabrice Bellard 已提交
411 412 413 414 415
/* generic function for encode/decode called after a frame has been coded/decoded */
void MPV_frame_end(MpegEncContext *s)
{
    /* draw edge for correct motion prediction if outside */
    if (s->pict_type != B_TYPE) {
416
      if(s->avctx->codec->id!=CODEC_ID_MPEG4){
417 418 419
        draw_edges(s->current_picture[0], s->linesize, s->mb_width*16, s->mb_height*16, EDGE_WIDTH);
        draw_edges(s->current_picture[1], s->linesize/2, s->mb_width*8, s->mb_height*8, EDGE_WIDTH/2);
        draw_edges(s->current_picture[2], s->linesize/2, s->mb_width*8, s->mb_height*8, EDGE_WIDTH/2);
420 421
      }else{
        /* OpenDivx, but i dunno how to distinguish it from mpeg4 */
F
Fabrice Bellard 已提交
422 423 424
        draw_edges(s->current_picture[0], s->linesize, s->width, s->height, EDGE_WIDTH);
        draw_edges(s->current_picture[1], s->linesize/2, s->width/2, s->height/2, EDGE_WIDTH/2);
        draw_edges(s->current_picture[2], s->linesize/2, s->width/2, s->height/2, EDGE_WIDTH/2);
425
      }
F
Fabrice Bellard 已提交
426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665
    }
}

int MPV_encode_picture(AVCodecContext *avctx,
                       unsigned char *buf, int buf_size, void *data)
{
    MpegEncContext *s = avctx->priv_data;
    AVPicture *pict = data;
    int i, j;

    if (s->fixed_qscale) 
        s->qscale = avctx->quality;

    init_put_bits(&s->pb, buf, buf_size, NULL, NULL);

    if (!s->intra_only) {
        /* first picture of GOP is intra */
        if ((s->picture_number % s->gop_size) == 0)
            s->pict_type = I_TYPE;
        else
            s->pict_type = P_TYPE;
    } else {
        s->pict_type = I_TYPE;
    }
    avctx->key_frame = (s->pict_type == I_TYPE);
    
    MPV_frame_start(s);

    for(i=0;i<3;i++) {
        UINT8 *src = pict->data[i];
        UINT8 *dest = s->current_picture[i];
        int src_wrap = pict->linesize[i];
        int dest_wrap = s->linesize;
        int w = s->width;
        int h = s->height;

        if (i >= 1) {
            dest_wrap >>= 1;
            w >>= 1;
            h >>= 1;
        }

        for(j=0;j<h;j++) {
            memcpy(dest, src, w);
            dest += dest_wrap;
            src += src_wrap;
        }
        s->new_picture[i] = s->current_picture[i];
    }

    encode_picture(s, s->picture_number);
    
    MPV_frame_end(s);
    s->picture_number++;

    if (s->out_format == FMT_MJPEG)
        mjpeg_picture_trailer(s);

    flush_put_bits(&s->pb);
    s->total_bits += (s->pb.buf_ptr - s->pb.buf) * 8;
    avctx->quality = s->qscale;
    return s->pb.buf_ptr - s->pb.buf;
}

static inline int clip(int a, int amin, int amax)
{
    if (a < amin)
        return amin;
    else if (a > amax)
        return amax;
    else
        return a;
}

/* apply one mpeg motion vector to the three components */
static inline void mpeg_motion(MpegEncContext *s,
                               UINT8 *dest_y, UINT8 *dest_cb, UINT8 *dest_cr,
                               int dest_offset,
                               UINT8 **ref_picture, int src_offset,
                               int field_based, op_pixels_func *pix_op,
                               int motion_x, int motion_y, int h)
{
    UINT8 *ptr;
    int dxy, offset, mx, my, src_x, src_y, height, linesize;
    
    dxy = ((motion_y & 1) << 1) | (motion_x & 1);
    src_x = s->mb_x * 16 + (motion_x >> 1);
    src_y = s->mb_y * (16 >> field_based) + (motion_y >> 1);
                
    /* WARNING: do no forget half pels */
    height = s->height >> field_based;
    src_x = clip(src_x, -16, s->width);
    if (src_x == s->width)
        dxy &= ~1;
    src_y = clip(src_y, -16, height);
    if (src_y == height)
        dxy &= ~2;
    linesize = s->linesize << field_based;
    ptr = ref_picture[0] + (src_y * linesize) + (src_x) + src_offset;
    dest_y += dest_offset;
    pix_op[dxy](dest_y, ptr, linesize, h);
    pix_op[dxy](dest_y + 8, ptr + 8, linesize, h);

    if (s->out_format == FMT_H263) {
        dxy = 0;
        if ((motion_x & 3) != 0)
            dxy |= 1;
        if ((motion_y & 3) != 0)
            dxy |= 2;
        mx = motion_x >> 2;
        my = motion_y >> 2;
    } else {
        mx = motion_x / 2;
        my = motion_y / 2;
        dxy = ((my & 1) << 1) | (mx & 1);
        mx >>= 1;
        my >>= 1;
    }
    
    src_x = s->mb_x * 8 + mx;
    src_y = s->mb_y * (8 >> field_based) + my;
    src_x = clip(src_x, -8, s->width >> 1);
    if (src_x == (s->width >> 1))
        dxy &= ~1;
    src_y = clip(src_y, -8, height >> 1);
    if (src_y == (height >> 1))
        dxy &= ~2;

    offset = (src_y * (linesize >> 1)) + src_x + (src_offset >> 1);
    ptr = ref_picture[1] + offset;
    pix_op[dxy](dest_cb + (dest_offset >> 1), ptr, linesize >> 1, h >> 1);
    ptr = ref_picture[2] + offset;
    pix_op[dxy](dest_cr + (dest_offset >> 1), ptr, linesize >> 1, h >> 1);
}

static inline void MPV_motion(MpegEncContext *s, 
                              UINT8 *dest_y, UINT8 *dest_cb, UINT8 *dest_cr,
                              int dir, UINT8 **ref_picture, 
                              op_pixels_func *pix_op)
{
    int dxy, offset, mx, my, src_x, src_y, motion_x, motion_y;
    int mb_x, mb_y, i;
    UINT8 *ptr, *dest;

    mb_x = s->mb_x;
    mb_y = s->mb_y;

    switch(s->mv_type) {
    case MV_TYPE_16X16:
        mpeg_motion(s, dest_y, dest_cb, dest_cr, 0,
                    ref_picture, 0,
                    0, pix_op,
                    s->mv[dir][0][0], s->mv[dir][0][1], 16);
        break;
    case MV_TYPE_8X8:
        for(i=0;i<4;i++) {
            motion_x = s->mv[dir][i][0];
            motion_y = s->mv[dir][i][1];

            dxy = ((motion_y & 1) << 1) | (motion_x & 1);
            src_x = mb_x * 16 + (motion_x >> 1) + (i & 1) * 8;
            src_y = mb_y * 16 + (motion_y >> 1) + ((i >> 1) & 1) * 8;
                    
            /* WARNING: do no forget half pels */
            src_x = clip(src_x, -16, s->width);
            if (src_x == s->width)
                dxy &= ~1;
            src_y = clip(src_y, -16, s->height);
            if (src_y == s->height)
                dxy &= ~2;
                    
            ptr = ref_picture[0] + (src_y * s->linesize) + (src_x);
            dest = dest_y + ((i & 1) * 8) + (i >> 1) * 8 * s->linesize;
            pix_op[dxy](dest, ptr, s->linesize, 8);
        }
        /* In case of 8X8, we construct a single chroma motion vector
           with a special rounding */
        mx = 0;
        my = 0;
        for(i=0;i<4;i++) {
            mx += s->mv[dir][i][0];
            my += s->mv[dir][i][1];
        }
        if (mx >= 0)
            mx = (h263_chroma_roundtab[mx & 0xf] + ((mx >> 3) & ~1));
        else {
            mx = -mx;
            mx = -(h263_chroma_roundtab[mx & 0xf] + ((mx >> 3) & ~1));
        }
        if (my >= 0)
            my = (h263_chroma_roundtab[my & 0xf] + ((my >> 3) & ~1));
        else {
            my = -my;
            my = -(h263_chroma_roundtab[my & 0xf] + ((my >> 3) & ~1));
        }
        dxy = ((my & 1) << 1) | (mx & 1);
        mx >>= 1;
        my >>= 1;

        src_x = mb_x * 8 + mx;
        src_y = mb_y * 8 + my;
        src_x = clip(src_x, -8, s->width/2);
        if (src_x == s->width/2)
            dxy &= ~1;
        src_y = clip(src_y, -8, s->height/2);
        if (src_y == s->height/2)
            dxy &= ~2;
        
        offset = (src_y * (s->linesize >> 1)) + src_x;
        ptr = ref_picture[1] + offset;
        pix_op[dxy](dest_cb, ptr, s->linesize >> 1, 8);
        ptr = ref_picture[2] + offset;
        pix_op[dxy](dest_cr, ptr, s->linesize >> 1, 8);
        break;
    case MV_TYPE_FIELD:
        if (s->picture_structure == PICT_FRAME) {
            /* top field */
            mpeg_motion(s, dest_y, dest_cb, dest_cr, 0,
                        ref_picture, s->field_select[dir][0] ? s->linesize : 0,
                        1, pix_op,
                        s->mv[dir][0][0], s->mv[dir][0][1], 8);
            /* bottom field */
            mpeg_motion(s, dest_y, dest_cb, dest_cr, s->linesize,
                        ref_picture, s->field_select[dir][1] ? s->linesize : 0,
                        1, pix_op,
                        s->mv[dir][1][0], s->mv[dir][1][1], 8);
        } else {
            

        }
        break;
    }
}


/* put block[] to dest[] */
static inline void put_dct(MpegEncContext *s, 
                           DCTELEM *block, int i, UINT8 *dest, int line_size)
{
    if (!s->mpeg2)
666
        s->dct_unquantize(s, block, i, s->qscale);
667
    ff_idct (block);
F
Fabrice Bellard 已提交
668 669 670 671 672 673 674 675 676
    put_pixels_clamped(block, dest, line_size);
}

/* add block[] to dest[] */
static inline void add_dct(MpegEncContext *s, 
                           DCTELEM *block, int i, UINT8 *dest, int line_size)
{
    if (s->block_last_index[i] >= 0) {
        if (!s->mpeg2)
677
            s->dct_unquantize(s, block, i, s->qscale);
678
        ff_idct (block);
F
Fabrice Bellard 已提交
679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701
        add_pixels_clamped(block, dest, line_size);
    }
}

/* generic function called after a macroblock has been parsed by the
   decoder or after it has been encoded by the encoder.

   Important variables used:
   s->mb_intra : true if intra macroblock
   s->mv_dir   : motion vector direction
   s->mv_type  : motion vector type
   s->mv       : motion vector
   s->interlaced_dct : true if interlaced dct used (mpeg2)
 */
void MPV_decode_mb(MpegEncContext *s, DCTELEM block[6][64])
{
    int mb_x, mb_y, motion_x, motion_y;
    int dct_linesize, dct_offset;
    op_pixels_func *op_pix;

    mb_x = s->mb_x;
    mb_y = s->mb_y;

702 703 704 705 706
#ifdef FF_POSTPROCESS
    quant_store[mb_y][mb_x]=s->qscale;
    //printf("[%02d][%02d] %d\n",mb_x,mb_y,s->qscale);
#endif

F
Fabrice Bellard 已提交
707 708 709
    /* update DC predictors for P macroblocks */
    if (!s->mb_intra) {
        if (s->h263_pred) {
710 711
          if(s->mbintra_table[mb_x + mb_y*s->mb_width])
          {
F
Fabrice Bellard 已提交
712
            int wrap, x, y, v;
713 714
            s->mbintra_table[mb_x + mb_y*s->mb_width]=0;
  
F
Fabrice Bellard 已提交
715 716 717 718
            wrap = 2 * s->mb_width + 2;
            v = 1024;
            x = 2 * mb_x + 1;
            y = 2 * mb_y + 1;
719
            
F
Fabrice Bellard 已提交
720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743
            s->dc_val[0][(x) + (y) * wrap] = v;
            s->dc_val[0][(x + 1) + (y) * wrap] = v;
            s->dc_val[0][(x) + (y + 1) * wrap] = v;
            s->dc_val[0][(x + 1) + (y + 1) * wrap] = v;
            /* ac pred */
            memset(s->ac_val[0][(x) + (y) * wrap], 0, 16 * sizeof(INT16));
            memset(s->ac_val[0][(x + 1) + (y) * wrap], 0, 16 * sizeof(INT16));
            memset(s->ac_val[0][(x) + (y + 1) * wrap], 0, 16 * sizeof(INT16));
            memset(s->ac_val[0][(x + 1) + (y + 1) * wrap], 0, 16 * sizeof(INT16));
            if (s->h263_msmpeg4) {
                s->coded_block[(x) + (y) * wrap] = 0;
                s->coded_block[(x + 1) + (y) * wrap] = 0;
                s->coded_block[(x) + (y + 1) * wrap] = 0;
                s->coded_block[(x + 1) + (y + 1) * wrap] = 0;
            }
            /* chroma */
            wrap = s->mb_width + 2;
            x = mb_x + 1;
            y = mb_y + 1;
            s->dc_val[1][(x) + (y) * wrap] = v;
            s->dc_val[2][(x) + (y) * wrap] = v;
            /* ac pred */
            memset(s->ac_val[1][(x) + (y) * wrap], 0, 16 * sizeof(INT16));
            memset(s->ac_val[2][(x) + (y) * wrap], 0, 16 * sizeof(INT16));
744
          }
F
Fabrice Bellard 已提交
745 746 747 748 749 750
        } else {
            s->last_dc[0] = 128 << s->intra_dc_precision;
            s->last_dc[1] = 128 << s->intra_dc_precision;
            s->last_dc[2] = 128 << s->intra_dc_precision;
        }
    }
751 752 753
    else
        s->mbintra_table[mb_x + mb_y*s->mb_width]=1;

F
Fabrice Bellard 已提交
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
    /* update motion predictor */
    if (s->out_format == FMT_H263) {
        int x, y, wrap;
        
        x = 2 * mb_x + 1;
        y = 2 * mb_y + 1;
        wrap = 2 * s->mb_width + 2;
        if (s->mb_intra) {
            motion_x = 0;
            motion_y = 0;
            goto motion_init;
        } else if (s->mv_type == MV_TYPE_16X16) {
            motion_x = s->mv[0][0][0];
            motion_y = s->mv[0][0][1];
        motion_init:
            /* no update if 8X8 because it has been done during parsing */
            s->motion_val[(x) + (y) * wrap][0] = motion_x;
            s->motion_val[(x) + (y) * wrap][1] = motion_y;
            s->motion_val[(x + 1) + (y) * wrap][0] = motion_x;
            s->motion_val[(x + 1) + (y) * wrap][1] = motion_y;
            s->motion_val[(x) + (y + 1) * wrap][0] = motion_x;
            s->motion_val[(x) + (y + 1) * wrap][1] = motion_y;
            s->motion_val[(x + 1) + (y + 1) * wrap][0] = motion_x;
            s->motion_val[(x + 1) + (y + 1) * wrap][1] = motion_y;
        }
    }
    
    if (!s->intra_only) {
        UINT8 *dest_y, *dest_cb, *dest_cr;
783 784 785
        UINT8 *mbskip_ptr;

        /* avoid copy if macroblock skipped in last frame too */
786
        if (!s->encoding && s->pict_type != B_TYPE) {
787 788 789 790 791 792 793 794 795 796 797
            mbskip_ptr = &s->mbskip_table[s->mb_y * s->mb_width + s->mb_x];
            if (s->mb_skiped) {
                s->mb_skiped = 0;
                /* if previous was skipped too, then nothing to do ! */
                if (*mbskip_ptr != 0) 
                    goto the_end;
                *mbskip_ptr = 1; /* indicate that this time we skiped it */
            } else {
                *mbskip_ptr = 0; /* not skipped */
            }
        }
F
Fabrice Bellard 已提交
798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834

        dest_y = s->current_picture[0] + (mb_y * 16 * s->linesize) + mb_x * 16;
        dest_cb = s->current_picture[1] + (mb_y * 8 * (s->linesize >> 1)) + mb_x * 8;
        dest_cr = s->current_picture[2] + (mb_y * 8 * (s->linesize >> 1)) + mb_x * 8;

        if (s->interlaced_dct) {
            dct_linesize = s->linesize * 2;
            dct_offset = s->linesize;
        } else {
            dct_linesize = s->linesize;
            dct_offset = s->linesize * 8;
        }

        if (!s->mb_intra) {
            /* motion handling */
            if (!s->no_rounding) 
                op_pix = put_pixels_tab;
            else
                op_pix = put_no_rnd_pixels_tab;

            if (s->mv_dir & MV_DIR_FORWARD) {
                MPV_motion(s, dest_y, dest_cb, dest_cr, 0, s->last_picture, op_pix);
                if (!s->no_rounding) 
                    op_pix = avg_pixels_tab;
                else
                    op_pix = avg_no_rnd_pixels_tab;
            }
            if (s->mv_dir & MV_DIR_BACKWARD) {
                MPV_motion(s, dest_y, dest_cb, dest_cr, 1, s->next_picture, op_pix);
            }

            /* add dct residue */
            add_dct(s, block[0], 0, dest_y, dct_linesize);
            add_dct(s, block[1], 1, dest_y + 8, dct_linesize);
            add_dct(s, block[2], 2, dest_y + dct_offset, dct_linesize);
            add_dct(s, block[3], 3, dest_y + dct_offset + 8, dct_linesize);

F
Fabrice Bellard 已提交
835 836
            add_dct(s, block[4], 4, dest_cb, s->linesize >> 1);
            add_dct(s, block[5], 5, dest_cr, s->linesize >> 1);
F
Fabrice Bellard 已提交
837 838 839 840 841 842 843
        } else {
            /* dct only in intra block */
            put_dct(s, block[0], 0, dest_y, dct_linesize);
            put_dct(s, block[1], 1, dest_y + 8, dct_linesize);
            put_dct(s, block[2], 2, dest_y + dct_offset, dct_linesize);
            put_dct(s, block[3], 3, dest_y + dct_offset + 8, dct_linesize);

F
Fabrice Bellard 已提交
844 845
            put_dct(s, block[4], 4, dest_cb, s->linesize >> 1);
            put_dct(s, block[5], 5, dest_cr, s->linesize >> 1);
F
Fabrice Bellard 已提交
846 847
        }
    }
848 849
 the_end:
    emms_c();
F
Fabrice Bellard 已提交
850 851 852 853
}

static void encode_picture(MpegEncContext *s, int picture_number)
{
854
    int mb_x, mb_y, wrap, last_gob;
F
Fabrice Bellard 已提交
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
    UINT8 *ptr;
    int i, motion_x, motion_y;

    s->picture_number = picture_number;
    if (!s->fixed_qscale) 
        s->qscale = rate_estimate_qscale(s);

    /* precompute matrix */
    if (s->out_format == FMT_MJPEG) {
        /* for mjpeg, we do include qscale in the matrix */
        s->intra_matrix[0] = default_intra_matrix[0];
        for(i=1;i<64;i++)
            s->intra_matrix[i] = (default_intra_matrix[i] * s->qscale) >> 3;
        convert_matrix(s->q_intra_matrix, s->intra_matrix, 8);
    } else {
        convert_matrix(s->q_intra_matrix, s->intra_matrix, s->qscale);
        convert_matrix(s->q_non_intra_matrix, s->non_intra_matrix, s->qscale);
    }

    switch(s->out_format) {
    case FMT_MJPEG:
        mjpeg_picture_header(s);
        break;
    case FMT_H263:
        if (s->h263_msmpeg4) 
            msmpeg4_encode_picture_header(s, picture_number);
        else if (s->h263_pred)
            mpeg4_encode_picture_header(s, picture_number);
        else if (s->h263_rv10) 
            rv10_encode_picture_header(s, picture_number);
        else
            h263_encode_picture_header(s, picture_number);
        break;
    case FMT_MPEG1:
        mpeg1_encode_picture_header(s, picture_number);
        break;
    }
        
    /* init last dc values */
    /* note: quant matrix value (8) is implied here */
    s->last_dc[0] = 128;
    s->last_dc[1] = 128;
    s->last_dc[2] = 128;
    s->mb_incr = 1;
    s->last_mv[0][0][0] = 0;
    s->last_mv[0][0][1] = 0;
    s->mv_type = MV_TYPE_16X16;
    s->mv_dir = MV_DIR_FORWARD;

904 905 906 907 908 909 910 911 912 913
    /* Get the GOB height based on picture height */
    if (s->out_format == FMT_H263 && s->h263_plus) {
        if (s->height <= 400)
            s->gob_index = 1;
        else if (s->height <= 800)
            s->gob_index = 2;
        else
            s->gob_index = 4;
    }
        
F
Fabrice Bellard 已提交
914
    for(mb_y=0; mb_y < s->mb_height; mb_y++) {
915 916 917 918 919 920 921 922 923 924 925 926
        /* Put GOB header based on RTP MTU */
        if (!mb_y) {
            s->ptr_lastgob = s->pb.buf_ptr;
            s->ptr_last_mb_line = s->pb.buf_ptr;
        } else if (s->out_format == FMT_H263 && s->h263_plus) {
            last_gob = h263_encode_gob_header(s, mb_y);
            if (last_gob) {
                //fprintf(stderr,"\nLast GOB size: %d", last_gob);
                s->first_gob_line = 1;
            } else
                s->first_gob_line = 0;
        }
F
Fabrice Bellard 已提交
927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945
        for(mb_x=0; mb_x < s->mb_width; mb_x++) {

            s->mb_x = mb_x;
            s->mb_y = mb_y;

            /* compute motion vector and macro block type (intra or non intra) */
            motion_x = 0;
            motion_y = 0;
            if (s->pict_type == P_TYPE) {
                s->mb_intra = estimate_motion(s, mb_x, mb_y,
                                              &motion_x,
                                              &motion_y);
            } else {
                s->mb_intra = 1;
            }

            /* get the pixels */
            wrap = s->linesize;
            ptr = s->new_picture[0] + (mb_y * 16 * wrap) + mb_x * 16;
946 947 948 949
            get_pixels(s->block[0], ptr, wrap);
            get_pixels(s->block[1], ptr + 8, wrap);
            get_pixels(s->block[2], ptr + 8 * wrap, wrap);
            get_pixels(s->block[3], ptr + 8 * wrap + 8, wrap);
F
Fabrice Bellard 已提交
950 951
            wrap = s->linesize >> 1;
            ptr = s->new_picture[1] + (mb_y * 8 * wrap) + mb_x * 8;
952
            get_pixels(s->block[4], ptr, wrap);
F
Fabrice Bellard 已提交
953 954 955

            wrap = s->linesize >> 1;
            ptr = s->new_picture[2] + (mb_y * 8 * wrap) + mb_x * 8;
956
            get_pixels(s->block[5], ptr, wrap);
F
Fabrice Bellard 已提交
957 958 959 960 961 962 963 964 965 966

            /* subtract previous frame if non intra */
            if (!s->mb_intra) {
                int dxy, offset, mx, my;

                dxy = ((motion_y & 1) << 1) | (motion_x & 1);
                ptr = s->last_picture[0] + 
                    ((mb_y * 16 + (motion_y >> 1)) * s->linesize) + 
                    (mb_x * 16 + (motion_x >> 1));

967 968 969 970
                sub_pixels_2(s->block[0], ptr, s->linesize, dxy);
                sub_pixels_2(s->block[1], ptr + 8, s->linesize, dxy);
                sub_pixels_2(s->block[2], ptr + s->linesize * 8, s->linesize, dxy);
                sub_pixels_2(s->block[3], ptr + 8 + s->linesize * 8, s->linesize ,dxy);
F
Fabrice Bellard 已提交
971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989

                if (s->out_format == FMT_H263) {
                    /* special rounding for h263 */
                    dxy = 0;
                    if ((motion_x & 3) != 0)
                        dxy |= 1;
                    if ((motion_y & 3) != 0)
                        dxy |= 2;
                    mx = motion_x >> 2;
                    my = motion_y >> 2;
                } else {
                    mx = motion_x / 2;
                    my = motion_y / 2;
                    dxy = ((my & 1) << 1) | (mx & 1);
                    mx >>= 1;
                    my >>= 1;
                }
                offset = ((mb_y * 8 + my) * (s->linesize >> 1)) + (mb_x * 8 + mx);
                ptr = s->last_picture[1] + offset;
990
                sub_pixels_2(s->block[4], ptr, s->linesize >> 1, dxy);
F
Fabrice Bellard 已提交
991
                ptr = s->last_picture[2] + offset;
992
                sub_pixels_2(s->block[5], ptr, s->linesize >> 1, dxy);
F
Fabrice Bellard 已提交
993
            }
994
            emms_c();
F
Fabrice Bellard 已提交
995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009

            /* DCT & quantize */
            if (s->h263_msmpeg4) {
                msmpeg4_dc_scale(s);
            } else if (s->h263_pred) {
                h263_dc_scale(s);
            } else {
                /* default quantization values */
                s->y_dc_scale = 8;
                s->c_dc_scale = 8;
            }

            for(i=0;i<6;i++) {
                int last_index;
                if (av_fdct == jpeg_fdct_ifast)
1010
                    last_index = dct_quantize(s, s->block[i], i, s->qscale);
F
Fabrice Bellard 已提交
1011
                else
1012
                    last_index = dct_quantize_mmx(s, s->block[i], i, s->qscale);
F
Fabrice Bellard 已提交
1013 1014 1015 1016 1017 1018
                s->block_last_index[i] = last_index;
            }

            /* huffman encode */
            switch(s->out_format) {
            case FMT_MPEG1:
1019
                mpeg1_encode_mb(s, s->block, motion_x, motion_y);
F
Fabrice Bellard 已提交
1020 1021 1022
                break;
            case FMT_H263:
                if (s->h263_msmpeg4)
1023
                    msmpeg4_encode_mb(s, s->block, motion_x, motion_y);
F
Fabrice Bellard 已提交
1024
                else
1025
                    h263_encode_mb(s, s->block, motion_x, motion_y);
F
Fabrice Bellard 已提交
1026 1027
                break;
            case FMT_MJPEG:
1028
                mjpeg_encode_mb(s, s->block);
F
Fabrice Bellard 已提交
1029 1030 1031 1032 1033 1034 1035
                break;
            }

            /* decompress blocks so that we keep the state of the decoder */
            s->mv[0][0][0] = motion_x;
            s->mv[0][0][1] = motion_y;

1036
            MPV_decode_mb(s, s->block);
F
Fabrice Bellard 已提交
1037
        }
1038 1039 1040 1041 1042 1043 1044 1045
        /* Obtain average MB line size for RTP */
        if (!mb_y)
            s->mb_line_avgsize = s->pb.buf_ptr - s->ptr_last_mb_line;
        else    
            s->mb_line_avgsize = (s->mb_line_avgsize + s->pb.buf_ptr - s->ptr_last_mb_line) >> 1;
        //fprintf(stderr, "\nMB line: %d\tSize: %u\tAvg. Size: %u", s->mb_y, 
        //                    (s->pb.buf_ptr - s->ptr_last_mb_line), s->mb_line_avgsize);
        s->ptr_last_mb_line = s->pb.buf_ptr;
F
Fabrice Bellard 已提交
1046
    }
1047 1048
    //if (s->gob_number)
    //    fprintf(stderr,"\nNumber of GOB: %d", s->gob_number);
F
Fabrice Bellard 已提交
1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059
}

static int dct_quantize(MpegEncContext *s, 
                        DCTELEM *block, int n,
                        int qscale)
{
    int i, j, level, last_non_zero, q;
    const int *qmat;

    av_fdct (block);

F
Fabrice Bellard 已提交
1060 1061 1062 1063
    /* we need this permutation so that we correct the IDCT
       permutation. will be moved into DCT code */
    block_permute(block);

F
Fabrice Bellard 已提交
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
    if (s->mb_intra) {
        if (n < 4)
            q = s->y_dc_scale;
        else
            q = s->c_dc_scale;
        q = q << 3;
        
        /* note: block[0] is assumed to be positive */
        block[0] = (block[0] + (q >> 1)) / q;
        i = 1;
        last_non_zero = 0;
        if (s->out_format == FMT_H263) {
            qmat = s->q_non_intra_matrix;
        } else {
            qmat = s->q_intra_matrix;
        }
    } else {
        i = 0;
        last_non_zero = -1;
        qmat = s->q_non_intra_matrix;
    }

    for(;i<64;i++) {
        j = zigzag_direct[i];
        level = block[j];
        level = level * qmat[j];
#ifdef PARANOID
        {
            static int count = 0;
            int level1, level2, qmat1;
            double val;
            if (qmat == s->q_non_intra_matrix) {
                qmat1 = default_non_intra_matrix[j] * s->qscale;
            } else {
                qmat1 = default_intra_matrix[j] * s->qscale;
            }
            if (av_fdct != jpeg_fdct_ifast)
                val = ((double)block[j] * 8.0) / (double)qmat1;
            else
                val = ((double)block[j] * 8.0 * 2048.0) / 
                    ((double)qmat1 * aanscales[j]);
            level1 = (int)val;
            level2 = level / (1 << (QMAT_SHIFT - 3));
            if (level1 != level2) {
                fprintf(stderr, "%d: quant error qlevel=%d wanted=%d level=%d qmat1=%d qmat=%d wantedf=%0.6f\n", 
                        count, level2, level1, block[j], qmat1, qmat[j],
                        val);
                count++;
            }

        }
#endif
        /* XXX: slight error for the low range. Test should be equivalent to
           (level <= -(1 << (QMAT_SHIFT - 3)) || level >= (1 <<
           (QMAT_SHIFT - 3)))
        */
        if (((level << (31 - (QMAT_SHIFT - 3))) >> (31 - (QMAT_SHIFT - 3))) != 
            level) {
            level = level / (1 << (QMAT_SHIFT - 3));
            /* XXX: currently, this code is not optimal. the range should be:
               mpeg1: -255..255
               mpeg2: -2048..2047
               h263:  -128..127
               mpeg4: -2048..2047
            */
            if (level > 127)
                level = 127;
            else if (level < -128)
                level = -128;
            block[j] = level;
            last_non_zero = i;
        } else {
            block[j] = 0;
        }
    }
    return last_non_zero;
}

static int dct_quantize_mmx(MpegEncContext *s, 
                            DCTELEM *block, int n,
                            int qscale)
{
    int i, j, level, last_non_zero, q;
    const int *qmat;

    av_fdct (block);
1150 1151 1152 1153
    
    /* we need this permutation so that we correct the IDCT
       permutation. will be moved into DCT code */
    block_permute(block);
F
Fabrice Bellard 已提交
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

    if (s->mb_intra) {
        if (n < 4)
            q = s->y_dc_scale;
        else
            q = s->c_dc_scale;
        
        /* note: block[0] is assumed to be positive */
        block[0] = (block[0] + (q >> 1)) / q;
        i = 1;
        last_non_zero = 0;
        if (s->out_format == FMT_H263) {
            qmat = s->q_non_intra_matrix;
        } else {
            qmat = s->q_intra_matrix;
        }
    } else {
        i = 0;
        last_non_zero = -1;
        qmat = s->q_non_intra_matrix;
    }

    for(;i<64;i++) {
        j = zigzag_direct[i];
        level = block[j];
        level = level * qmat[j];
        /* XXX: slight error for the low range. Test should be equivalent to
           (level <= -(1 << (QMAT_SHIFT_MMX - 3)) || level >= (1 <<
           (QMAT_SHIFT_MMX - 3)))
        */
        if (((level << (31 - (QMAT_SHIFT_MMX - 3))) >> (31 - (QMAT_SHIFT_MMX - 3))) != 
            level) {
            level = level / (1 << (QMAT_SHIFT_MMX - 3));
            /* XXX: currently, this code is not optimal. the range should be:
               mpeg1: -255..255
               mpeg2: -2048..2047
               h263:  -128..127
               mpeg4: -2048..2047
            */
            if (level > 127)
                level = 127;
            else if (level < -128)
                level = -128;
            block[j] = level;
            last_non_zero = i;
        } else {
            block[j] = 0;
        }
    }
    return last_non_zero;
}

1206 1207
static void dct_unquantize_mpeg1_c(MpegEncContext *s, 
                                   DCTELEM *block, int n, int qscale)
F
Fabrice Bellard 已提交
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
{
    int i, level;
    const UINT16 *quant_matrix;

    if (s->mb_intra) {
        if (n < 4) 
            block[0] = block[0] * s->y_dc_scale;
        else
            block[0] = block[0] * s->c_dc_scale;
        /* XXX: only mpeg1 */
        quant_matrix = s->intra_matrix;
        for(i=1;i<64;i++) {
            level = block[i];
            if (level) {
                if (level < 0) {
                    level = -level;
                    level = (int)(level * qscale * quant_matrix[i]) >> 3;
                    level = (level - 1) | 1;
                    level = -level;
                } else {
                    level = (int)(level * qscale * quant_matrix[i]) >> 3;
                    level = (level - 1) | 1;
                }
#ifdef PARANOID
                if (level < -2048 || level > 2047)
                    fprintf(stderr, "unquant error %d %d\n", i, level);
#endif
                block[i] = level;
            }
        }
    } else {
        i = 0;
        quant_matrix = s->non_intra_matrix;
        for(;i<64;i++) {
            level = block[i];
            if (level) {
                if (level < 0) {
                    level = -level;
                    level = (((level << 1) + 1) * qscale *
                             ((int) (quant_matrix[i]))) >> 4;
                    level = (level - 1) | 1;
                    level = -level;
                } else {
                    level = (((level << 1) + 1) * qscale *
                             ((int) (quant_matrix[i]))) >> 4;
                    level = (level - 1) | 1;
                }
#ifdef PARANOID
                if (level < -2048 || level > 2047)
                    fprintf(stderr, "unquant error %d %d\n", i, level);
#endif
                block[i] = level;
            }
        }
    }
}
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

static void dct_unquantize_h263_c(MpegEncContext *s, 
                                  DCTELEM *block, int n, int qscale)
{
    int i, level, qmul, qadd;

    if (s->mb_intra) {
        if (n < 4) 
            block[0] = block[0] * s->y_dc_scale;
        else
            block[0] = block[0] * s->c_dc_scale;
        i = 1;
    } else {
        i = 0;
    }

    qmul = s->qscale << 1;
    qadd = (s->qscale - 1) | 1;

    for(;i<64;i++) {
        level = block[i];
        if (level) {
            if (level < 0) {
                level = level * qmul - qadd;
            } else {
                level = level * qmul + qadd;
            }
#ifdef PARANOID
                if (level < -2048 || level > 2047)
                    fprintf(stderr, "unquant error %d %d\n", i, level);
#endif
            block[i] = level;
        }
    }
}
F
Fabrice Bellard 已提交
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

/* rate control */

/* an I frame is I_FRAME_SIZE_RATIO bigger than a P frame */
#define I_FRAME_SIZE_RATIO 3.0
#define QSCALE_K           20

static void rate_control_init(MpegEncContext *s)
{
    s->wanted_bits = 0;

    if (s->intra_only) {
        s->I_frame_bits = ((INT64)s->bit_rate * FRAME_RATE_BASE) / s->frame_rate;
        s->P_frame_bits = s->I_frame_bits;
    } else {
        s->P_frame_bits = (int) ((float)(s->gop_size * s->bit_rate) / 
                                 (float)((float)s->frame_rate / FRAME_RATE_BASE * (I_FRAME_SIZE_RATIO + s->gop_size - 1)));
        s->I_frame_bits = (int)(s->P_frame_bits * I_FRAME_SIZE_RATIO);
    }
    
#if defined(DEBUG)
    printf("I_frame_size=%d P_frame_size=%d\n",
           s->I_frame_bits, s->P_frame_bits);
#endif
}


/*
 * This heuristic is rather poor, but at least we do not have to
 * change the qscale at every macroblock.
 */
static int rate_estimate_qscale(MpegEncContext *s)
{
1332
    INT64 diff, total_bits = s->total_bits;
F
Fabrice Bellard 已提交
1333
    float q;
1334
    int qscale, qmin;
F
Fabrice Bellard 已提交
1335 1336 1337 1338 1339 1340 1341 1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356

    if (s->pict_type == I_TYPE) {
        s->wanted_bits += s->I_frame_bits;
    } else {
        s->wanted_bits += s->P_frame_bits;
    }
    diff = s->wanted_bits - total_bits;
    q = 31.0 - (float)diff / (QSCALE_K * s->mb_height * s->mb_width);
    /* adjust for I frame */
    if (s->pict_type == I_TYPE && !s->intra_only) {
        q /= I_FRAME_SIZE_RATIO;
    }

    /* using a too small Q scale leeds to problems in mpeg1 and h263
       because AC coefficients are clamped to 255 or 127 */
    qmin = 3;
    if (q < qmin)
        q = qmin;
    else if (q > 31)
        q = 31;
    qscale = (int)(q + 0.5);
#if defined(DEBUG)
F
Fabrice Bellard 已提交
1357
    printf("%d: total=%0.0f br=%0.1f diff=%d qest=%0.1f\n", 
F
Fabrice Bellard 已提交
1358
           s->picture_number, 
F
Fabrice Bellard 已提交
1359
           (double)total_bits, 
F
Fabrice Bellard 已提交
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 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416
           (float)s->frame_rate / FRAME_RATE_BASE * 
           total_bits / s->picture_number, 
           diff, q);
#endif
    return qscale;
}

AVCodec mpeg1video_encoder = {
    "mpeg1video",
    CODEC_TYPE_VIDEO,
    CODEC_ID_MPEG1VIDEO,
    sizeof(MpegEncContext),
    MPV_encode_init,
    MPV_encode_picture,
    MPV_encode_end,
};

AVCodec h263_encoder = {
    "h263",
    CODEC_TYPE_VIDEO,
    CODEC_ID_H263,
    sizeof(MpegEncContext),
    MPV_encode_init,
    MPV_encode_picture,
    MPV_encode_end,
};

AVCodec h263p_encoder = {
    "h263p",
    CODEC_TYPE_VIDEO,
    CODEC_ID_H263P,
    sizeof(MpegEncContext),
    MPV_encode_init,
    MPV_encode_picture,
    MPV_encode_end,
};

AVCodec rv10_encoder = {
    "rv10",
    CODEC_TYPE_VIDEO,
    CODEC_ID_RV10,
    sizeof(MpegEncContext),
    MPV_encode_init,
    MPV_encode_picture,
    MPV_encode_end,
};

AVCodec mjpeg_encoder = {
    "mjpeg",
    CODEC_TYPE_VIDEO,
    CODEC_ID_MJPEG,
    sizeof(MpegEncContext),
    MPV_encode_init,
    MPV_encode_picture,
    MPV_encode_end,
};

F
Fabrice Bellard 已提交
1417 1418
AVCodec mpeg4_encoder = {
    "mpeg4",
F
Fabrice Bellard 已提交
1419
    CODEC_TYPE_VIDEO,
F
Fabrice Bellard 已提交
1420
    CODEC_ID_MPEG4,
F
Fabrice Bellard 已提交
1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435
    sizeof(MpegEncContext),
    MPV_encode_init,
    MPV_encode_picture,
    MPV_encode_end,
};

AVCodec msmpeg4_encoder = {
    "msmpeg4",
    CODEC_TYPE_VIDEO,
    CODEC_ID_MSMPEG4,
    sizeof(MpegEncContext),
    MPV_encode_init,
    MPV_encode_picture,
    MPV_encode_end,
};