frame.c

另一个版本的x264的decoder

static inline void deblocking_filter_edgev( x264_t *h, uint8_t *pix, int i_pix_stride, int bS[4], int i_QP )
{
    int i, d;
    const int i_index_a = x264_clip3( i_QP + h->sh.i_alpha_c0_offset, 0, 51 );
    const int alpha = i_alpha_table[i_index_a];
    const int beta  = i_beta_table[x264_clip3( i_QP + h->sh.i_beta_offset, 0, 51 )];

    for( i = 0; i < 4; i++ )
    {
        if( bS[i] == 0 )
        {
            pix += 4 * i_pix_stride;
            continue;
        }

        if( bS[i] < 4 )
        {
            const int tc0 = i_tc0_table[i_index_a][bS[i] - 1];

            /* 4px edge length */
            for( d = 0; d < 4; d++ )
            {
                const int p0 = pix[-1];
                const int p1 = pix[-2];
                const int p2 = pix[-3];
                const int q0 = pix[0];
                const int q1 = pix[1];
                const int q2 = pix[2];

                if( abs( p0 - q0 ) < alpha &&
                    abs( p1 - p0 ) < beta &&
                    abs( q1 - q0 ) < beta )
                {
                    int tc = tc0;
                    int i_delta;

                    if( abs( p2 - p0 ) < beta )
                    {
                        pix[-2] = p1 + x264_clip3( ( p2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( p1 << 1 ) ) >> 1, -tc0, tc0 );
                        tc++;
                    }
                    if( abs( q2 - q0 ) < beta )
                    {
                        pix[1] = q1 + x264_clip3( ( q2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( q1 << 1 ) ) >> 1, -tc0, tc0 );
                        tc++;
                    }

                    i_delta = x264_clip3( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
                    pix[-1] = clip_uint8( p0 + i_delta );    /* p0' */
                    pix[0]  = clip_uint8( q0 - i_delta );    /* q0' */
                }
                pix += i_pix_stride;
            }
        }
        else
        {
            /* 4px edge length */
            for( d = 0; d < 4; d++ )
            {
                const int p0 = pix[-1];
                const int p1 = pix[-2];
                const int p2 = pix[-3];

                const int q0 = pix[0];
                const int q1 = pix[1];
                const int q2 = pix[2];

                if( abs( p0 - q0 ) < alpha &&
                    abs( p1 - p0 ) < beta &&
                    abs( q1 - q0 ) < beta )
                {
                    if( abs( p0 - q0 ) < (( alpha >> 2 ) + 2 ) )
                    {
                        if( abs( p2 - p0 ) < beta )
                        {
                            const int p3 = pix[-4];
                            /* p0', p1', p2' */
                            pix[-1] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
                            pix[-2] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
                            pix[-3] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
                        }
                        else
                        {
                            /* p0' */
                            pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
                        }
                        if( abs( q2 - q0 ) < beta )
                        {
                            const int q3 = pix[3];
                            /* q0', q1', q2' */
                            pix[0] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
                            pix[1] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
                            pix[2] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
                        }
                        else
                        {
                            /* q0' */
                            pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
                        }
                    }
                    else
                    {
                        /* p0', q0' */
                        pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
                        pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
                    }
                }
                pix += i_pix_stride;
            }
        }
    }
}

static inline void deblocking_filter_edgecv( x264_t *h, uint8_t *pix, int i_pix_stride, int bS[4], int i_QP )
{
    int i, d;
    const int i_index_a = x264_clip3( i_QP + h->sh.i_alpha_c0_offset, 0, 51 );
    const int alpha = i_alpha_table[i_index_a];
    const int beta  = i_beta_table[x264_clip3( i_QP + h->sh.i_beta_offset, 0, 51 )];

    for( i = 0; i < 4; i++ )
    {
        if( bS[i] == 0 )
        {
            pix += 2 * i_pix_stride;
            continue;
        }

        if( bS[i] < 4 )
        {
            const int tc = i_tc0_table[i_index_a][bS[i] - 1] + 1;
            /* 2px edge length (because we use same bS than the one for luma) */
            for( d = 0; d < 2; d++ )
            {
                const int p0 = pix[-1];
                const int p1 = pix[-2];
                const int q0 = pix[0];
                const int q1 = pix[1];

                if( abs( p0 - q0 ) < alpha &&
                    abs( p1 - p0 ) < beta &&
                    abs( q1 - q0 ) < beta )
                {
                    const int i_delta = x264_clip3( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );

                    pix[-1] = clip_uint8( p0 + i_delta );    /* p0' */
                    pix[0]  = clip_uint8( q0 - i_delta );    /* q0' */
                }
                pix += i_pix_stride;
            }
        }
        else
        {
            /* 2px edge length (because we use same bS than the one for luma) */
            for( d = 0; d < 2; d++ )
            {
                const int p0 = pix[-1];
                const int p1 = pix[-2];
                const int q0 = pix[0];
                const int q1 = pix[1];

                if( abs( p0 - q0 ) < alpha &&
                    abs( p1 - p0 ) < beta &&
                    abs( q1 - q0 ) < beta )
                {
                    pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2;   /* p0' */
                    pix[0]  = ( 2*q1 + q0 + p1 + 2 ) >> 2;   /* q0' */
                }
                pix += i_pix_stride;
            }
        }
    }
}

static inline void deblocking_filter_edgeh( x264_t *h, uint8_t *pix, int i_pix_stride, int bS[4], int i_QP )
{
    int i, d;
    const int i_index_a = x264_clip3( i_QP + h->sh.i_alpha_c0_offset, 0, 51 );
    const int alpha = i_alpha_table[i_index_a];
    const int beta  = i_beta_table[x264_clip3( i_QP + h->sh.i_beta_offset, 0, 51 )];

    int i_pix_next  = i_pix_stride;

    for( i = 0; i < 4; i++ )
    {
        if( bS[i] == 0 )
        {
            pix += 4;
            continue;
        }

        if( bS[i] < 4 )
        {
            const int tc0 = i_tc0_table[i_index_a][bS[i] - 1];
            /* 4px edge length */
            for( d = 0; d < 4; d++ )
            {
                const int p0 = pix[-i_pix_next];
                const int p1 = pix[-2*i_pix_next];
                const int p2 = pix[-3*i_pix_next];
                const int q0 = pix[0];
                const int q1 = pix[1*i_pix_next];
                const int q2 = pix[2*i_pix_next];

                if( abs( p0 - q0 ) < alpha &&
                    abs( p1 - p0 ) < beta &&
                    abs( q1 - q0 ) < beta )
                {
                    int tc = tc0;
                    int i_delta;

                    if( abs( p2 - p0 ) < beta )
                    {
                        pix[-2*i_pix_next] = p1 + x264_clip3( ( p2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( p1 << 1 ) ) >> 1, -tc0, tc0 );
                        tc++;
                    }
                    if( abs( q2 - q0 ) < beta )
                    {
                        pix[i_pix_next] = q1 + x264_clip3( ( q2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( q1 << 1 ) ) >> 1, -tc0, tc0 );
                        tc++;
                    }

                    i_delta = x264_clip3( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
                    pix[-i_pix_next] = clip_uint8( p0 + i_delta );    /* p0' */
                    pix[0]           = clip_uint8( q0 - i_delta );    /* q0' */
                }
                pix++;
            }
        }
        else
        {
            /* 4px edge length */
            for( d = 0; d < 4; d++ )
            {
                const int p0 = pix[-i_pix_next];
                const int p1 = pix[-2*i_pix_next];
                const int p2 = pix[-3*i_pix_next];
                const int q0 = pix[0];
                const int q1 = pix[1*i_pix_next];
                const int q2 = pix[2*i_pix_next];

                if( abs( p0 - q0 ) < alpha &&
                    abs( p1 - p0 ) < beta &&
                    abs( q1 - q0 ) < beta )
                {
                    const int p3 = pix[-4*i_pix_next];
                    const int q3 = pix[ 3*i_pix_next];

                    if( abs( p0 - q0 ) < (( alpha >> 2 ) + 2 ) )
                    {
                        if( abs( p2 - p0 ) < beta )
                        {
                            /* p0', p1', p2' */
                            pix[-1*i_pix_next] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
                            pix[-2*i_pix_next] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
                            pix[-3*i_pix_next] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
                        }
                        else
                        {
                            /* p0' */
                            pix[-1*i_pix_next] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
                        }
                        if( abs( q2 - q0 ) < beta )
                        {
                            /* q0', q1', q2' */
                            pix[0*i_pix_next] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
                            pix[1*i_pix_next] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
                            pix[2*i_pix_next] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
                        }
                        else
                        {
                            /* q0' */
                            pix[0*i_pix_next] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
                        }
                    }
                    else
                    {
                        /* p0' */
                        pix[-1*i_pix_next] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
                        /* q0' */
                        pix[0*i_pix_next] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
                    }
                }
                pix++;