frame.c.svn-base

X264最新源码

/*****************************************************************************
 * frame.c: h264 encoder library
 *****************************************************************************
 * Copyright (C) 2003 Laurent Aimar
 * $Id: frame.c,v 1.1 2004/06/03 19:27:06 fenrir Exp $
 *
 * Authors: Laurent Aimar <[e-mail]>
 *
 * 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., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
 *****************************************************************************/

#include "common.h"

#define PADH 32
#define PADV 32

x264_frame_t *x264_frame_new( x264_t *h )
{
    x264_frame_t *frame = x264_malloc( sizeof(x264_frame_t) );
    int i, j;

    int i_mb_count = h->mb.i_mb_count;
    int i_stride;
    int i_lines;
    int i_padv = PADV << h->param.b_interlaced;

    if( !frame ) return NULL;

    memset( frame, 0, sizeof(x264_frame_t) );

    /* allocate frame data (+64 for extra data for me) */
    i_stride = ( ( h->param.i_width  + 15 ) & -16 )+ 2*PADH;
    i_lines  = ( ( h->param.i_height + 15 ) & -16 );
    if( h->param.b_interlaced )
        i_lines = ( i_lines + 31 ) & -32;

    frame->i_plane = 3;
    for( i = 0; i < 3; i++ )
    {
        int i_divh = 1;
        int i_divw = 1;
        if( i > 0 )
        {
            if( h->param.i_csp == X264_CSP_I420 )
                i_divh = i_divw = 2;
            else if( h->param.i_csp == X264_CSP_I422 )
                i_divw = 2;
        }
        frame->i_stride[i] = i_stride / i_divw;
        frame->i_lines[i] = i_lines / i_divh;
        CHECKED_MALLOC( frame->buffer[i],
                        frame->i_stride[i] * ( frame->i_lines[i] + 2*i_padv / i_divh ) );

        frame->plane[i] = ((uint8_t*)frame->buffer[i]) +
                          frame->i_stride[i] * i_padv / i_divh + PADH / i_divw;
    }
    frame->i_stride[3] = 0;
    frame->i_lines[3] = 0;
    frame->buffer[3] = NULL;
    frame->plane[3] = NULL;

    frame->filtered[0] = frame->plane[0];
    for( i = 0; i < 3; i++ )
    {
        CHECKED_MALLOC( frame->buffer[4+i],
                        frame->i_stride[0] * ( frame->i_lines[0] + 2*i_padv ) );
        frame->filtered[i+1] = ((uint8_t*)frame->buffer[4+i]) +
                                frame->i_stride[0] * i_padv + PADH;
    }

    if( h->frames.b_have_lowres )
    {
        frame->i_stride_lowres = frame->i_stride[0]/2 + PADH;
        frame->i_lines_lowres = frame->i_lines[0]/2;
        for( i = 0; i < 4; i++ )
        {
            CHECKED_MALLOC( frame->buffer_lowres[i],
                            frame->i_stride_lowres * ( frame->i_lines[0]/2 + 2*i_padv ) );
            frame->lowres[i] = ((uint8_t*)frame->buffer_lowres[i]) +
                                frame->i_stride_lowres * i_padv + PADH;
        }
    }

    if( h->param.analyse.i_me_method == X264_ME_ESA )
    {
        CHECKED_MALLOC( frame->buffer[7],
                        2 * frame->i_stride[0] * (frame->i_lines[0] + 2*i_padv) * sizeof(uint16_t) );
        frame->integral = (uint16_t*)frame->buffer[7] + frame->i_stride[0] * i_padv + PADH;
    }

    frame->i_poc = -1;
    frame->i_type = X264_TYPE_AUTO;
    frame->i_qpplus1 = 0;
    frame->i_pts = -1;
    frame->i_frame = -1;
    frame->i_frame_num = -1;
    frame->i_lines_completed = -1;

    CHECKED_MALLOC( frame->mb_type, i_mb_count * sizeof(int8_t));
    CHECKED_MALLOC( frame->mv[0], 2*16 * i_mb_count * sizeof(int16_t) );
    CHECKED_MALLOC( frame->ref[0], 4 * i_mb_count * sizeof(int8_t) );
    if( h->param.i_bframe )
    {
        CHECKED_MALLOC( frame->mv[1], 2*16 * i_mb_count * sizeof(int16_t) );
        CHECKED_MALLOC( frame->ref[1], 4 * i_mb_count * sizeof(int8_t) );
    }
    else
    {
        frame->mv[1]  = NULL;
        frame->ref[1] = NULL;
    }

    CHECKED_MALLOC( frame->i_row_bits, i_lines/16 * sizeof(int) );
    CHECKED_MALLOC( frame->i_row_qp, i_lines/16 * sizeof(int) );
    for( i = 0; i < h->param.i_bframe + 2; i++ )
        for( j = 0; j < h->param.i_bframe + 2; j++ )
            CHECKED_MALLOC( frame->i_row_satds[i][j], i_lines/16 * sizeof(int) );

    pthread_mutex_init( &frame->mutex, NULL );
    pthread_cond_init( &frame->cv, NULL );

    return frame;

fail:
    x264_frame_delete( frame );
    return NULL;
}

void x264_frame_delete( x264_frame_t *frame )
{
    int i, j;
    for( i = 0; i < 8; i++ )
        x264_free( frame->buffer[i] );
    for( i = 0; i < 4; i++ )
        x264_free( frame->buffer_lowres[i] );
    for( i = 0; i < X264_BFRAME_MAX+2; i++ )
        for( j = 0; j < X264_BFRAME_MAX+2; j++ )
            x264_free( frame->i_row_satds[i][j] );
    x264_free( frame->i_row_bits );
    x264_free( frame->i_row_qp );
    x264_free( frame->mb_type );
    x264_free( frame->mv[0] );
    x264_free( frame->mv[1] );
    x264_free( frame->ref[0] );
    x264_free( frame->ref[1] );
    pthread_mutex_destroy( &frame->mutex );
    pthread_cond_destroy( &frame->cv );
    x264_free( frame );
}

void x264_frame_copy_picture( x264_t *h, x264_frame_t *dst, x264_picture_t *src )
{
    int i_csp = src->img.i_csp & X264_CSP_MASK;
    dst->i_type     = src->i_type;
    dst->i_qpplus1  = src->i_qpplus1;
    dst->i_pts      = src->i_pts;

    if( i_csp <= X264_CSP_NONE  || i_csp >= X264_CSP_MAX )
        x264_log( h, X264_LOG_ERROR, "Arg invalid CSP\n" );
    else
        h->csp.convert[i_csp]( &h->mc, dst, &src->img, h->param.i_width, h->param.i_height );
}



static void plane_expand_border( uint8_t *pix, int i_stride, int i_width, int i_height, int i_padh, int i_padv, int b_pad_top, int b_pad_bottom )
{
#define PPIXEL(x, y) ( pix + (x) + (y)*i_stride )
    int y;
    for( y = 0; y < i_height; y++ )
    {
        /* left band */
        memset( PPIXEL(-i_padh, y), PPIXEL(0, y)[0], i_padh );
        /* right band */
        memset( PPIXEL(i_width, y), PPIXEL(i_width-1, y)[0], i_padh );
    }
    /* upper band */
    if( b_pad_top )
    for( y = 0; y < i_padv; y++ )
        memcpy( PPIXEL(-i_padh, -y-1), PPIXEL(-i_padh, 0), i_width+2*i_padh );
    /* lower band */
    if( b_pad_bottom )
    for( y = 0; y < i_padv; y++ )
        memcpy( PPIXEL(-i_padh, i_height+y), PPIXEL(-i_padh, i_height-1), i_width+2*i_padh );
#undef PPIXEL
}

void x264_frame_expand_border( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
{
    int i;
    int b_start = !mb_y;
    if( mb_y & h->sh.b_mbaff )
        return;
    for( i = 0; i < frame->i_plane; i++ )
    {
        int stride = frame->i_stride[i];
        int width = 16*h->sps->i_mb_width >> !!i;
        int height = (b_end ? 16*(h->sps->i_mb_height - mb_y) >> h->sh.b_mbaff : 16) >> !!i;
        int padh = PADH >> !!i;
        int padv = PADV >> !!i;
        // buffer: 2 chroma, 3 luma (rounded to 4) because deblocking goes beyond the top of the mb
        uint8_t *pix = frame->plane[i] + X264_MAX(0, (16*mb_y-4)*stride >> !!i);
        if( b_end && !b_start )
            height += 4 >> (!!i + h->sh.b_mbaff);
        if( h->sh.b_mbaff )
        {
            plane_expand_border( pix, stride*2, width, height, padh, padv, b_start, b_end );
            plane_expand_border( pix+stride, stride*2, width, height, padh, padv, b_start, b_end );
        }
        else
        {
            plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end );
        }
    }
}

void x264_frame_expand_border_filtered( x264_t *h, x264_frame_t *frame, int mb_y, int b_end )
{
    /* during filtering, 8 extra pixels were filtered on each edge. 
       we want to expand border from the last filtered pixel */
    int b_start = !mb_y;
    int stride = frame->i_stride[0];
    int width = 16*h->sps->i_mb_width + 16;
    int height = b_end ? (16*(h->sps->i_mb_height - mb_y) >> h->sh.b_mbaff) + 16 : 16;
    int padh = PADH - 8;
    int padv = PADV - 8;
    int i;
    for( i = 1; i < 4; i++ )
    {
        // buffer: 8 luma, to match the hpel filter
        uint8_t *pix = frame->filtered[i] + (16*mb_y - (8 << h->sh.b_mbaff)) * stride - 8;
        if( h->sh.b_mbaff )
        {
            plane_expand_border( pix, stride*2, width, height, padh, padv, b_start, b_end );
            plane_expand_border( pix+stride, stride*2, width, height, padh, padv, b_start, b_end );
        }
        else
        {
            plane_expand_border( pix, stride, width, height, padh, padv, b_start, b_end );
        }
    }
}

void x264_frame_expand_border_lowres( x264_frame_t *frame )
{
    int i;
    for( i = 0; i < 4; i++ )
        plane_expand_border( frame->lowres[i], frame->i_stride_lowres, frame->i_stride_lowres - 2*PADH, frame->i_lines_lowres, PADH, PADV, 1, 1 );
}

void x264_frame_expand_border_mod16( x264_t *h, x264_frame_t *frame )
{
    int i, y;
    for( i = 0; i < frame->i_plane; i++ )
    {
        int i_subsample = i ? 1 : 0;
        int i_width = h->param.i_width >> i_subsample;
        int i_height = h->param.i_height >> i_subsample;
        int i_padx = ( h->sps->i_mb_width * 16 - h->param.i_width ) >> i_subsample;
        int i_pady = ( h->sps->i_mb_height * 16 - h->param.i_height ) >> i_subsample;

        if( i_padx )
        {
            for( y = 0; y < i_height; y++ )
                memset( &frame->plane[i][y*frame->i_stride[i] + i_width],
                         frame->plane[i][y*frame->i_stride[i] + i_width - 1],
                         i_padx );
        }
        if( i_pady )
        {
            //FIXME interlace? or just let it pad using the wrong field
            for( y = i_height; y < i_height + i_pady; y++ )
                memcpy( &frame->plane[i][y*frame->i_stride[i]],
                        &frame->plane[i][(i_height-1)*frame->i_stride[i]],
                        i_width + i_padx );
        }
    }
}


/* cavlc + 8x8 transform stores nnz per 16 coeffs for the purpose of
 * entropy coding, but per 64 coeffs for the purpose of deblocking */
void munge_cavlc_nnz_row( x264_t *h, int mb_y, uint8_t (*buf)[16] )
{
    uint32_t (*src)[6] = (uint32_t(*)[6])h->mb.non_zero_count + mb_y * h->sps->i_mb_width;
    int8_t *transform = h->mb.mb_transform_size + mb_y * h->sps->i_mb_width;
    int x;
    for( x=0; x<h->sps->i_mb_width; x++ )
    {
        memcpy( buf+x, src+x, 16 );
        if( transform[x] )
        {
            if( src[x][0] ) src[x][0] = 0x01010101;