edge-detec.c.svn-base

H.264 source codes

        i_refb = -2;
        if( !IS_INTRA( mbn->i_type ) )
        {
            i_refb = mbn->partition[x][yn].i_ref[0];
            mvxb   = mbn->partition[x][yn].mv[0][0];
            mvyb   = mbn->partition[x][yn].mv[0][1];
        }
    }

    if( i_refa == -1 || i_refb == -1 ||
        ( i_refa == 0 && mvxa == 0 && mvya == 0 ) ||
        ( i_refb == 0 && mvxb == 0 && mvyb == 0 ) )
    {
        *mvxp = 0;
        *mvyp = 0;
    }
    else
    {
        x264_macroblock_predict_mv( mb, 0, 0, 0, mvxp, mvyp );
    }
}

static const int i_chroma_qp_table[52] =
{
     0,  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, 27, 28, 29,
    29, 30, 31, 32, 32, 33, 34, 34, 35, 35,
    36, 36, 37, 37, 37, 38, 38, 38, 39, 39,
    39, 39
};

static void x264_macroblock_mc( x264_t *h, x264_macroblock_t *mb, int b_luma )
{
    x264_mb_context_t *ctx = mb->context;

    int ch;
    int i_ref;
    int mx, my;

    if( mb->i_type == P_L0 )
    {
        int i_part;

        for( i_part = 0; i_part < mb_partition_count( mb->i_partition ); i_part++ )
        {
            int i_width, i_height;
            int x, y;

            x264_macroblock_partition_get( mb, 0, i_part, 0, &i_ref, &mx, &my );
            x264_macroblock_partition_getxy( mb, i_part, 0, &x, &y );
            x264_macroblock_partition_size(  mb, i_part, 0, &i_width, &i_height );

            if( b_luma )
            {
                int     i_src = ctx->i_fref0[i_ref][0];
                uint8_t *p_src= ctx->p_fref0[i_ref][0];
                int     i_dst = ctx->i_fdec[0];
                uint8_t *p_dst= ctx->p_fdec[0];

                h->mc[MC_LUMA]( &p_src[4*(x+y*i_src)], i_src,
                                &p_dst[4*(x+y*i_dst)], i_dst,
                                mx, my, 4*i_width, 4*i_height );
            }
            else
            {
                int     i_src,  i_dst;
                uint8_t *p_src, *p_dst;

                for( ch = 0; ch < 2; ch++ )
                {
                    i_src = ctx->i_fref0[i_ref][1+ch];
                    p_src = ctx->p_fref0[i_ref][1+ch];
                    i_dst = ctx->i_fdec[1+ch];
                    p_dst = ctx->p_fdec[1+ch];

                    h->mc[MC_CHROMA]( &p_src[2*(x+y*i_src)], i_src,
                                      &p_dst[2*(x+y*i_dst)], i_dst,
                                      mx, my, 2*i_width, 2*i_height );
                }
            }
        }
    }
    else if( mb->i_type == P_8x8 )
    {
        int i_part;

        for( i_part = 0; i_part < 4; i_part++ )
        {
            int i_sub;

            for( i_sub = 0; i_sub < mb_sub_partition_count( mb->i_sub_partition[i_part] ); i_sub++ )
            {
                int i_width, i_height;
                int x, y;

                x264_macroblock_partition_get(   mb, 0, i_part, i_sub, &i_ref, &mx, &my );
                x264_macroblock_partition_getxy( mb, i_part, i_sub, &x, &y );
                x264_macroblock_partition_size(  mb, i_part, i_sub, &i_width, &i_height );

                if( b_luma )
                {
                    int     i_src = ctx->i_fref0[i_ref][0];
                    uint8_t *p_src= ctx->p_fref0[i_ref][0];
                    int     i_dst = ctx->i_fdec[0];
                    uint8_t *p_dst= ctx->p_fdec[0];

                    h->mc[MC_LUMA]( &p_src[4*(x+y*i_src)], i_src,
                                    &p_dst[4*(x+y*i_dst)], i_dst,
                                    mx, my, 4*i_width, 4*i_height );
                }
                else
                {
                    int     i_src,  i_dst;
                    uint8_t *p_src, *p_dst;

                    for( ch = 0; ch < 2; ch++ )
                    {
                        i_src = ctx->i_fref0[i_ref][1+ch];
                        p_src = ctx->p_fref0[i_ref][1+ch];
                        i_dst = ctx->i_fdec[1+ch];
                        p_dst = ctx->p_fdec[1+ch];

                        h->mc[MC_CHROMA]( &p_src[2*(x+y*i_src)], i_src,
                                          &p_dst[2*(x+y*i_dst)], i_dst,
                                          mx, my, 2*i_width, 2*i_height );
                    }
                }
            }
        }
    }
}

/*****************************************************************************
 * x264_macroblock_neighbour_load:
 *****************************************************************************/
void x264_macroblock_context_load( x264_t *h, x264_macroblock_t *mb, x264_mb_context_t *context )
{
    int i;
    int x, y;
    x264_macroblock_t *a = NULL;
    x264_macroblock_t *b = NULL;

    if( mb->i_neighbour&MB_LEFT )
    {
        a = mb - 1;
    }
    if( mb->i_neighbour&MB_TOP )
    {
        b = mb - h->sps.i_mb_width;
    }
#define LOAD_PTR( dst, src ) \
    context->p_##dst[0] = (src)->plane[0] + 16 * ( mb->i_mb_x + mb->i_mb_y * (src)->i_stride[0] ); \
    context->p_##dst[1] = (src)->plane[1] +  8 * ( mb->i_mb_x + mb->i_mb_y * (src)->i_stride[1] ); \
    context->p_##dst[2] = (src)->plane[2] +  8 * ( mb->i_mb_x + mb->i_mb_y * (src)->i_stride[2] ); \
    context->i_##dst[0] = (src)->i_stride[0]; \
    context->i_##dst[1] = (src)->i_stride[1]; \
    context->i_##dst[2] = (src)->i_stride[2]

    LOAD_PTR( img,  h->picture );
    LOAD_PTR( fdec, h->fdec );
    for( i = 0; i < h->i_ref0; i++ )
    {
        LOAD_PTR( fref0[i], h->fref0[i] );
    }
    for( i = 0; i < h->i_ref1; i++ )
    {
        LOAD_PTR( fref1[i], h->fref1[i] );
    }
#undef LOAD_PTR

    for( y = 0; y < 4; y++ )
    {
        for( x = 0; x < 4; x++ )
        {
            int idx;
            int xa, yb;
            x264_macroblock_t *mba;
            x264_macroblock_t *mbb;

            idx = block_idx_xy[x][y];
            mba = mb;
            mbb = mb;

            xa = x - 1;
            if (xa < 0 )
            {
                xa += 4;
                mba = a;
            }
            /* up */
            yb = y - 1;
            if (yb < 0 )
            {
                yb += 4;
                mbb = b;
            }

            context->block[idx].mba = mba;
            context->block[idx].mbb = mbb;
            context->block[idx].bka = mba ? &mba->block[block_idx_xy[xa][y]] : NULL;
            context->block[idx].bkb = mbb ? &mbb->block[block_idx_xy[x][yb]] : NULL;

            if( x < 2 && y < 2 )
            {
                int ch;
                if( xa > 1 ) xa -= 2;   /* we have wrap but here step is 2 not 4 */
                if( yb > 1 ) yb -= 2;   /* idem */

                for( ch = 0; ch < 2; ch++ )
                {
                    context->block[16+4*ch+idx].mba = mba;
                    context->block[16+4*ch+idx].mbb = mbb;
                    context->block[16+4*ch+idx].bka = mba ? &mba->block[16+4*ch+block_idx_xy[xa][y]] : NULL;
                    context->block[16+4*ch+idx].bkb = mbb ? &mbb->block[16+4*ch+block_idx_xy[x][yb]] : NULL;
                }
            }
        }
    }

    mb->context = context;
}

/* (ref: JVT-B118)
 * x264_mb_decimate_score: given dct coeffs it returns a score to see if we could empty this dct coeffs
 * to 0 (low score means set it to null)
 * Used in inter macroblock (luma and chroma)
 *  luma: for a 8x8 block: if score < 4 -> null
 *        for the complete mb: if score < 6 -> null
 *  chroma: for the complete mb: if score < 7 -> null
 */
static int x264_mb_decimate_score( int *dct, int i_max )
{
    static const int i_ds_table[16] = { 3, 2, 2, 1, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 };

    int i_score = 0;
    int idx = i_max - 1;

    while( idx >= 0 && dct[idx] == 0 )
    {
        idx--;
    }

    while( idx >= 0 )
    {
        int i_run;

        if( abs( dct[idx--] ) > 1 )
        {
            return 9;
        }

        i_run = 0;
        while( idx >= 0 && dct[idx] == 0 )
        {
            idx--;
            i_run++;
        }
        i_score += i_ds_table[i_run];
    }

    return i_score;
}

static void x264_mb_encode_4x4( x264_t *h, x264_macroblock_t *mb, int idx, int i_qscale )
{
    x264_mb_context_t *ctx = mb->context;

    uint8_t *p_src = ctx->p_img[0] + 4 * block_idx_x[idx] + 4 * block_idx_y[idx] * ctx->i_img[0];
    int      i_src = ctx->i_img[0];
    uint8_t *p_dst = ctx->p_fdec[0] + 4 * block_idx_x[idx] + 4 * block_idx_y[idx] * ctx->i_fdec[0];
    int      i_dst = ctx->i_fdec[0];

    int16_t luma[4][4];
    int16_t dct4x4[4][4];

    /* we calculate diff */
    h->pixf.sub4x4( luma, p_src, i_src, p_dst, i_dst );

    /* calculate dct coeffs */
    h->dctf.dct4x4( dct4x4, luma );
    quant_4x4( dct4x4, i_qscale, 1 );

    scan_zigzag_4x4full( mb->block[idx].luma4x4, dct4x4 );

    /* output samples to fdec */
    dequant_4x4( dct4x4, i_qscale );
    h->dctf.idct4x4( luma, dct4x4 );

    /* put pixel to fdec */
    h->pixf.add4x4( p_dst, i_dst, luma );
}

static void x264_mb_encode_i16x16( x264_t *h, x264_macroblock_t *mb, int i_qscale )
{
    x264_mb_context_t *ctx = mb->context;

    uint8_t *p_src = ctx->p_img[0];
    int      i_src = ctx->i_img[0];
    uint8_t *p_dst = ctx->p_fdec[0];
    int      i_dst = ctx->i_fdec[0];

    int16_t luma[16][4][4];
    int16_t dct4x4[16+1][4][4];

    int i;

    /* calculate the diff */
    h->pixf.sub16x16( luma, p_src, i_src, p_dst, i_dst );

    /* calculate dct coeffs */
    for( i = 0; i < 16; i++ )
    {
        h->dctf.dct4x4( dct4x4[i+1], luma[i] );

        /* copy dc coeff */
        dct4x4[0][block_idx_y[i]][block_idx_x[i]] = dct4x4[1+i][0][0];

        quant_4x4( dct4x4[1+i], i_qscale, 1 );
        scan_zigzag_4x4( mb->block[i].residual_ac, dct4x4[1+i] );
    }

    h->dctf.dct4x4dc( dct4x4[0], dct4x4[0] );
    quant_4x4_dc( dct4x4[0], i_qscale, 1 );
    scan_zigzag_4x4full( mb->luma16x16_dc, dct4x4[0] );

    /* output samples to fdec */
    h->dctf.idct4x4dc( dct4x4[0], dct4x4[0] );
    dequant_4x4_dc( dct4x4[0], i_qscale );  /* XXX not inversed */

    /* calculate dct coeffs */
    for( i = 0; i < 16; i++ )
    {
        dequant_4x4( dct4x4[1+i], i_qscale );

        /* copy dc coeff */
        dct4x4[1+i][0][0] = dct4x4[0][block_idx_y[i]][block_idx_x[i]];

        h->dctf.idct4x4( luma[i], dct4x4[i+1] );
    }
    /* put pixels to fdec */
    h->pixf.add16x16( p_dst, i_dst, luma );
}

static void x264_mb_encode_8x8( x264_t *h, x264_macroblock_t *mb, int b_inter, int i_qscale )
{
    x264_mb_context_t *ctx = mb->context;

    uint8_t *p_src, *p_dst;
    int      i_src, i_dst;

    int i, ch;
    int i_decimate_score = 0;

    for( ch = 0; ch < 2; ch++ )
    {
        int16_t chroma[4][4][4];
        int16_t dct2x2[2][2];
        int16_t dct4x4[4][4][4];

        p_src = ctx->p_img[1+ch];
        i_src = ctx->i_img[1+ch];
        p_dst = ctx->p_fdec[1+ch];
        i_dst = ctx->i_fdec[1+ch];

        /* calculate the diff */
        h->pixf.sub8x8( chroma, p_src, i_src, p_dst, i_dst );

        /* calculate dct coeffs */
        for( i = 0; i < 4; i++ )
        {
            h->dctf.dct4x4( dct4x4[i], chroma[i] );

            /* copy dc coeff */
            dct2x2[block_idx_y[i]][block_idx_x[i]] = dct4x4[i][0][0];

            quant_4x4( dct4x4[i], i_qscale, 1 );
            scan_zigzag_4x4( mb->block[16+i+ch*4].residual_ac, dct4x4[i] );

            i_decimate_score += x264_mb_decimate_score( mb->block[16+i+ch*4].residual_ac, 15 );
        }

        h->dctf.dct2x2dc( dct2x2, dct2x2 );
        quant_2x2_dc( dct2x2, i_qscale, 1 );
        scan_zigzag_2x2_dc( mb->chroma_dc[ch], dct2x2 );

        if( i_decimate_score < 7 && b_inter )
        {
            /* Near null chroma 8x8 block so make it null (bits saving) */
            for( i = 0; i < 4; i++ )
            {
                int x, y;
                for( x = 0; x < 15; x++ )
                {
                    mb->block[16+i+ch*4].residual_ac[x] = 0;
                }
                for( x = 0; x < 4; x++ )
                {
                    for( y = 0; y < 4; y++ )
                    {
                        dct4x4[i][x][y] = 0;
                    }
                }
            }
        }

        /* output samples to fdec */
        h->dctf.idct2x2dc( dct2x2, dct2x2 );
        dequant_2x2_dc( dct2x2, i_qscale );  /* XXX not inversed */

        /* calculate dct coeffs */
        for( i = 0; i < 4; i++ )
        {
            dequant_4x4( dct4x4[i], i_qscale );

            /* copy dc coeff */
            dct4x4[i][0][0] = dct2x2[block_idx_y[i]][block_idx_x[i]];

            h->dctf.idct4x4( chroma[i], dct4x4[i] );
        }
        h->pixf.add8x8( p_dst, i_dst, chroma );
    }
}

static int x264_mb_pred_mode4x4_fix( int i_mode )
{
    if( i_mode == I_PRED_4x4_DC_LEFT || i_mode == I_PRED_4x4_DC_TOP || i_mode == I_PRED_4x4_DC_128 )
    {
        return I_PRED_4x4_DC;
    }
    return i_mode;
}
static int x264_mb_pred_mode16x16_fix( int i_mode )
{
    if( i_mode == I_PRED_16x16_DC_LEFT || i_mode == I_PRED_16x16_DC_TOP || i_mode == I_PRED_16x16_DC_128 )
    {
        return I_PRED_16x16_DC;
    }
    return i_mode;
}