summaryrefslogtreecommitdiffstats
path: root/libavcodec/h264_loopfilter.c
diff options
context:
space:
mode:
Diffstat (limited to 'libavcodec/h264_loopfilter.c')
-rw-r--r--libavcodec/h264_loopfilter.c224
1 files changed, 59 insertions, 165 deletions
diff --git a/libavcodec/h264_loopfilter.c b/libavcodec/h264_loopfilter.c
index 325fd3c..2e61a31 100644
--- a/libavcodec/h264_loopfilter.c
+++ b/libavcodec/h264_loopfilter.c
@@ -101,197 +101,92 @@ static const uint8_t tc0_table[52*3][4] = {
};
static void av_always_inline filter_mb_edgev( uint8_t *pix, int stride, int16_t bS[4], unsigned int qp, H264Context *h) {
- const int bit_depth = h->sps.bit_depth_luma;
- const int qp_bd_offset = 6*(bit_depth-8);
+ const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);
const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;
- const int alpha = alpha_table[index_a] << (bit_depth-8);
- const int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset] << (bit_depth-8);
+ const int alpha = alpha_table[index_a];
+ const int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset];
if (alpha ==0 || beta == 0) return;
if( bS[0] < 4 ) {
int8_t tc[4];
- tc[0] = tc0_table[index_a][bS[0]] << (bit_depth-8);
- tc[1] = tc0_table[index_a][bS[1]] << (bit_depth-8);
- tc[2] = tc0_table[index_a][bS[2]] << (bit_depth-8);
- tc[3] = tc0_table[index_a][bS[3]] << (bit_depth-8);
+ tc[0] = tc0_table[index_a][bS[0]];
+ tc[1] = tc0_table[index_a][bS[1]];
+ tc[2] = tc0_table[index_a][bS[2]];
+ tc[3] = tc0_table[index_a][bS[3]];
h->h264dsp.h264_h_loop_filter_luma(pix, stride, alpha, beta, tc);
} else {
h->h264dsp.h264_h_loop_filter_luma_intra(pix, stride, alpha, beta);
}
}
static void av_always_inline filter_mb_edgecv( uint8_t *pix, int stride, int16_t bS[4], unsigned int qp, H264Context *h ) {
- const int bit_depth = h->sps.bit_depth_luma;
- const int qp_bd_offset = 6*(bit_depth-8);
+ const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);
const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;
- const int alpha = alpha_table[index_a] << (bit_depth-8);
- const int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset] << (bit_depth-8);
+ const int alpha = alpha_table[index_a];
+ const int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset];
if (alpha ==0 || beta == 0) return;
if( bS[0] < 4 ) {
int8_t tc[4];
- tc[0] = (tc0_table[index_a][bS[0]] << (bit_depth-8))+1;
- tc[1] = (tc0_table[index_a][bS[1]] << (bit_depth-8))+1;
- tc[2] = (tc0_table[index_a][bS[2]] << (bit_depth-8))+1;
- tc[3] = (tc0_table[index_a][bS[3]] << (bit_depth-8))+1;
+ tc[0] = tc0_table[index_a][bS[0]]+1;
+ tc[1] = tc0_table[index_a][bS[1]]+1;
+ tc[2] = tc0_table[index_a][bS[2]]+1;
+ tc[3] = tc0_table[index_a][bS[3]]+1;
h->h264dsp.h264_h_loop_filter_chroma(pix, stride, alpha, beta, tc);
} else {
h->h264dsp.h264_h_loop_filter_chroma_intra(pix, stride, alpha, beta);
}
}
-static void filter_mb_mbaff_edgev( H264Context *h, uint8_t *pix, int stride, int16_t bS[4], int bsi, int qp ) {
- int i;
- const int bit_depth = h->sps.bit_depth_luma;
- const int qp_bd_offset = 6*(bit_depth-8);
+static void filter_mb_mbaff_edgev( H264Context *h, uint8_t *pix, int stride, int16_t bS[7], int bsi, int qp ) {
+ const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);
int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;
- int alpha = alpha_table[index_a] << (bit_depth-8);
- int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset] << (bit_depth-8);
- for( i = 0; i < 8; i++, pix += stride) {
- const int bS_index = (i >> 1) * bsi;
-
- if( bS[bS_index] == 0 ) {
- continue;
- }
-
- if( bS[bS_index] < 4 ) {
- const int tc0 = tc0_table[index_a][bS[bS_index]] << (bit_depth-8);
- 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( FFABS( p0 - q0 ) < alpha &&
- FFABS( p1 - p0 ) < beta &&
- FFABS( q1 - q0 ) < beta ) {
- int tc = tc0;
- int i_delta;
-
- if( FFABS( p2 - p0 ) < beta ) {
- if(tc0)
- pix[-2] = p1 + av_clip( ( p2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( p1 << 1 ) ) >> 1, -tc0, tc0 );
- tc++;
- }
- if( FFABS( q2 - q0 ) < beta ) {
- if(tc0)
- pix[1] = q1 + av_clip( ( q2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( q1 << 1 ) ) >> 1, -tc0, tc0 );
- tc++;
- }
-
- i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
- pix[-1] = av_clip_uint8( p0 + i_delta ); /* p0' */
- pix[0] = av_clip_uint8( q0 - i_delta ); /* q0' */
- tprintf(h->s.avctx, "filter_mb_mbaff_edgev i:%d, qp:%d, indexA:%d, alpha:%d, beta:%d, tc:%d\n# bS:%d -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x]\n", i, qp[qp_index], index_a, alpha, beta, tc, bS[bS_index], pix[-3], p1, p0, q0, q1, pix[2], p1, pix[-1], pix[0], q1);
- }
- }else{
- 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( FFABS( p0 - q0 ) < alpha &&
- FFABS( p1 - p0 ) < beta &&
- FFABS( q1 - q0 ) < beta ) {
+ int alpha = alpha_table[index_a];
+ int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset];
+ if (alpha ==0 || beta == 0) return;
- if(FFABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
- if( FFABS( 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( FFABS( 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;
- }
- tprintf(h->s.avctx, "filter_mb_mbaff_edgev i:%d, qp:%d, indexA:%d, alpha:%d, beta:%d\n# bS:4 -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x, %02x, %02x]\n", i, qp[qp_index], index_a, alpha, beta, p2, p1, p0, q0, q1, q2, pix[-3], pix[-2], pix[-1], pix[0], pix[1], pix[2]);
- }
- }
+ if( bS[0] < 4 ) {
+ int8_t tc[4];
+ tc[0] = tc0_table[index_a][bS[0*bsi]];
+ tc[1] = tc0_table[index_a][bS[1*bsi]];
+ tc[2] = tc0_table[index_a][bS[2*bsi]];
+ tc[3] = tc0_table[index_a][bS[3*bsi]];
+ h->h264dsp.h264_h_loop_filter_luma_mbaff(pix, stride, alpha, beta, tc);
+ } else {
+ h->h264dsp.h264_h_loop_filter_luma_mbaff_intra(pix, stride, alpha, beta);
}
}
-static void filter_mb_mbaff_edgecv( H264Context *h, uint8_t *pix, int stride, int16_t bS[4], int bsi, int qp ) {
- int i;
- const int bit_depth = h->sps.bit_depth_luma;
- const int qp_bd_offset = 6*(bit_depth-8);
+static void filter_mb_mbaff_edgecv( H264Context *h, uint8_t *pix, int stride, int16_t bS[7], int bsi, int qp ) {
+ const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);
int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;
- int alpha = alpha_table[index_a] << (bit_depth-8);
- int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset] << (bit_depth-8);
- for( i = 0; i < 4; i++, pix += stride) {
- const int bS_index = i*bsi;
-
- if( bS[bS_index] == 0 ) {
- continue;
- }
+ int alpha = alpha_table[index_a];
+ int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset];
+ if (alpha ==0 || beta == 0) return;
- if( bS[bS_index] < 4 ) {
- const int tc = (tc0_table[index_a][bS[bS_index]] << (bit_depth-8)) + 1;
- const int p0 = pix[-1];
- const int p1 = pix[-2];
- const int q0 = pix[0];
- const int q1 = pix[1];
-
- if( FFABS( p0 - q0 ) < alpha &&
- FFABS( p1 - p0 ) < beta &&
- FFABS( q1 - q0 ) < beta ) {
- const int i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
-
- pix[-1] = av_clip_uint8( p0 + i_delta ); /* p0' */
- pix[0] = av_clip_uint8( q0 - i_delta ); /* q0' */
- tprintf(h->s.avctx, "filter_mb_mbaff_edgecv i:%d, qp:%d, indexA:%d, alpha:%d, beta:%d, tc:%d\n# bS:%d -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x]\n", i, qp[qp_index], index_a, alpha, beta, tc, bS[bS_index], pix[-3], p1, p0, q0, q1, pix[2], p1, pix[-1], pix[0], q1);
- }
- }else{
- const int p0 = pix[-1];
- const int p1 = pix[-2];
- const int q0 = pix[0];
- const int q1 = pix[1];
-
- if( FFABS( p0 - q0 ) < alpha &&
- FFABS( p1 - p0 ) < beta &&
- FFABS( q1 - q0 ) < beta ) {
-
- pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2; /* p0' */
- pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2; /* q0' */
- tprintf(h->s.avctx, "filter_mb_mbaff_edgecv i:%d\n# bS:4 -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x, %02x, %02x]\n", i, pix[-3], p1, p0, q0, q1, pix[2], pix[-3], pix[-2], pix[-1], pix[0], pix[1], pix[2]);
- }
- }
+ if( bS[0] < 4 ) {
+ int8_t tc[4];
+ tc[0] = tc0_table[index_a][bS[0*bsi]] + 1;
+ tc[1] = tc0_table[index_a][bS[1*bsi]] + 1;
+ tc[2] = tc0_table[index_a][bS[2*bsi]] + 1;
+ tc[3] = tc0_table[index_a][bS[3*bsi]] + 1;
+ h->h264dsp.h264_h_loop_filter_chroma_mbaff(pix, stride, alpha, beta, tc);
+ } else {
+ h->h264dsp.h264_h_loop_filter_chroma_mbaff_intra(pix, stride, alpha, beta);
}
}
static void av_always_inline filter_mb_edgeh( uint8_t *pix, int stride, int16_t bS[4], unsigned int qp, H264Context *h ) {
- const int bit_depth = h->sps.bit_depth_luma;
- const int qp_bd_offset = 6*(bit_depth-8);
+ const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);
const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;
- const int alpha = alpha_table[index_a] << (bit_depth-8);
- const int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset] << (bit_depth-8);
+ const int alpha = alpha_table[index_a];
+ const int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset];
if (alpha ==0 || beta == 0) return;
if( bS[0] < 4 ) {
int8_t tc[4];
- tc[0] = tc0_table[index_a][bS[0]] << (bit_depth-8);
- tc[1] = tc0_table[index_a][bS[1]] << (bit_depth-8);
- tc[2] = tc0_table[index_a][bS[2]] << (bit_depth-8);
- tc[3] = tc0_table[index_a][bS[3]] << (bit_depth-8);
+ tc[0] = tc0_table[index_a][bS[0]];
+ tc[1] = tc0_table[index_a][bS[1]];
+ tc[2] = tc0_table[index_a][bS[2]];
+ tc[3] = tc0_table[index_a][bS[3]];
h->h264dsp.h264_v_loop_filter_luma(pix, stride, alpha, beta, tc);
} else {
h->h264dsp.h264_v_loop_filter_luma_intra(pix, stride, alpha, beta);
@@ -299,19 +194,18 @@ static void av_always_inline filter_mb_edgeh( uint8_t *pix, int stride, int16_t
}
static void av_always_inline filter_mb_edgech( uint8_t *pix, int stride, int16_t bS[4], unsigned int qp, H264Context *h ) {
- const int bit_depth = h->sps.bit_depth_luma;
- const int qp_bd_offset = 6*(bit_depth-8);
+ const int qp_bd_offset = 6 * (h->sps.bit_depth_luma - 8);
const unsigned int index_a = qp - qp_bd_offset + h->slice_alpha_c0_offset;
- const int alpha = alpha_table[index_a] << (bit_depth-8);
- const int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset] << (bit_depth-8);
+ const int alpha = alpha_table[index_a];
+ const int beta = beta_table[qp - qp_bd_offset + h->slice_beta_offset];
if (alpha ==0 || beta == 0) return;
if( bS[0] < 4 ) {
int8_t tc[4];
- tc[0] = (tc0_table[index_a][bS[0]] << (bit_depth-8))+1;
- tc[1] = (tc0_table[index_a][bS[1]] << (bit_depth-8))+1;
- tc[2] = (tc0_table[index_a][bS[2]] << (bit_depth-8))+1;
- tc[3] = (tc0_table[index_a][bS[3]] << (bit_depth-8))+1;
+ tc[0] = tc0_table[index_a][bS[0]]+1;
+ tc[1] = tc0_table[index_a][bS[1]]+1;
+ tc[2] = tc0_table[index_a][bS[2]]+1;
+ tc[3] = tc0_table[index_a][bS[3]]+1;
h->h264dsp.h264_v_loop_filter_chroma(pix, stride, alpha, beta, tc);
} else {
h->h264dsp.h264_v_loop_filter_chroma_intra(pix, stride, alpha, beta);
@@ -650,10 +544,10 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u
tprintf(s->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize);
//{ int i; for (i = 0; i < 4; i++) tprintf(s->avctx, " bS[%d]:%d", i, bS[i]); tprintf(s->avctx, "\n"); }
if( dir == 0 ) {
- filter_mb_edgev( &img_y[4*edge<<h->pixel_shift], linesize, bS, qp, h );
+ filter_mb_edgev( &img_y[4*edge << h->pixel_shift], linesize, bS, qp, h );
if( (edge&1) == 0 ) {
- filter_mb_edgecv( &img_cb[2*edge<<h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], h);
- filter_mb_edgecv( &img_cr[2*edge<<h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], h);
+ filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], h);
+ filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], h);
}
} else {
filter_mb_edgeh( &img_y[4*edge*linesize], linesize, bS, qp, h );
OpenPOWER on IntegriCloud