Mercurial > libavcodec.hg

comparison vc1.c @ 3520:a48c957d0377 libavcodec

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
Add rounding for overlap filters
author kostya
date Fri, 28 Jul 2006 03:27:37 +0000
parents 20938be7b67b
children ae749c2da968
comparison
equal deleted inserted replaced
3519:5ef29058581e 3520:a48c957d0377
829 829
830 /** Apply overlap transform to vertical edge 830 /** Apply overlap transform to vertical edge
831 * @todo optimize 831 * @todo optimize
832 * @todo move to DSPContext 832 * @todo move to DSPContext
833 */ 833 */
834 static void vc1_v_overlap(uint8_t* src, int stride) 834 static void vc1_v_overlap(uint8_t* src, int stride, int rnd)
835 { 835 {
836 int i; 836 int i;
837 int a, b, c, d; 837 int a, b, c, d;
838 for(i = 0; i < 8; i++) { 838 for(i = 0; i < 8; i++) {
839 a = src[-2*stride]; 839 a = src[-2*stride];
840 b = src[-stride]; 840 b = src[-stride];
841 c = src[0]; 841 c = src[0];
842 d = src[stride]; 842 d = src[stride];
843 843
844 src[-2*stride] = clip_uint8((7*a + d + 3) >> 3); 844 src[-2*stride] = clip_uint8((7*a + d + 4 - rnd) >> 3);
845 src[-stride] = clip_uint8((-a + 7*b + c + d + 3) >> 3); 845 src[-stride] = clip_uint8((-a + 7*b + c + d + 3 + rnd) >> 3);
846 src[0] = clip_uint8((a + b + 7*c - d + 3) >> 3); 846 src[0] = clip_uint8((a + b + 7*c - d + 4 - rnd) >> 3);
847 src[stride] = clip_uint8((a + 7*d + 3) >> 3); 847 src[stride] = clip_uint8((a + 7*d + 3 + rnd) >> 3);
848 src++; 848 src++;
849 } 849 }
850 } 850 }
851 851
852 /** Apply overlap transform to horizontal edge 852 /** Apply overlap transform to horizontal edge
853 * @todo optimize 853 * @todo optimize
854 * @todo move to DSPContext 854 * @todo move to DSPContext
855 */ 855 */
856 static void vc1_h_overlap(uint8_t* src, int stride) 856 static void vc1_h_overlap(uint8_t* src, int stride, int rnd)
857 { 857 {
858 int i; 858 int i;
859 int a, b, c, d; 859 int a, b, c, d;
860 for(i = 0; i < 8; i++) { 860 for(i = 0; i < 8; i++) {
861 a = src[-2]; 861 a = src[-2];
862 b = src[-1]; 862 b = src[-1];
863 c = src[0]; 863 c = src[0];
864 d = src[1]; 864 d = src[1];
865 865
866 src[-2] = clip_uint8((7*a + d + 3) >> 3); 866 src[-2] = clip_uint8((7*a + d + 4 - rnd) >> 3);
867 src[-1] = clip_uint8((-a + 7*b + c + d + 3) >> 3); 867 src[-1] = clip_uint8((-a + 7*b + c + d + 3 + rnd) >> 3);
868 src[0] = clip_uint8((a + b + 7*c - d + 3) >> 3); 868 src[0] = clip_uint8((a + b + 7*c - d + 4 - rnd) >> 3);
869 src[1] = clip_uint8((a + 7*d + 3) >> 3); 869 src[1] = clip_uint8((a + 7*d + 3 + rnd) >> 3);
870 src += stride; 870 src += stride;
871 } 871 }
872 } 872 }
873 873
874 /** Put block onto picture 874 /** Put block onto picture
2569 for(j = 0; j < 64; j++) s->block[i][j] += 128; 2569 for(j = 0; j < 64; j++) s->block[i][j] += 128;
2570 s->dsp.put_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2)); 2570 s->dsp.put_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
2571 /* TODO: proper loop filtering */ 2571 /* TODO: proper loop filtering */
2572 if(v->pq >= 9 && v->overlap) { 2572 if(v->pq >= 9 && v->overlap) {
2573 if(v->a_avail) 2573 if(v->a_avail)
2574 vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2)); 2574 vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2), (i<4) ? ((i&1)>>1) : (s->mb_y&1));
2575 if(v->c_avail) 2575 if(v->c_avail)
2576 vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2)); 2576 vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2), (i<4) ? (i&1) : (s->mb_x&1));
2577 } 2577 }
2578 } else if(val) { 2578 } else if(val) {
2579 vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block); 2579 vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block);
2580 if(!v->ttmbf && ttmb < 8) ttmb = -1; 2580 if(!v->ttmbf && ttmb < 8) ttmb = -1;
2581 first_block = 0; 2581 first_block = 0;
2670 for(j = 0; j < 64; j++) s->block[i][j] += 128; 2670 for(j = 0; j < 64; j++) s->block[i][j] += 128;
2671 s->dsp.put_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize); 2671 s->dsp.put_pixels_clamped(s->block[i], s->dest[dst_idx] + off, (i&4)?s->uvlinesize:s->linesize);
2672 /* TODO: proper loop filtering */ 2672 /* TODO: proper loop filtering */
2673 if(v->pq >= 9 && v->overlap) { 2673 if(v->pq >= 9 && v->overlap) {
2674 if(v->a_avail) 2674 if(v->a_avail)
2675 vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2)); 2675 vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2), (i<4) ? ((i&1)>>1) : (s->mb_y&1));
2676 if(v->c_avail) 2676 if(v->c_avail)
2677 vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2)); 2677 vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2), (i<4) ? (i&1) : (s->mb_x&1));
2678 } 2678 }
2679 } else if(is_coded[i]) { 2679 } else if(is_coded[i]) {
2680 status = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block); 2680 status = vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block);
2681 if(!v->ttmbf && ttmb < 8) ttmb = -1; 2681 if(!v->ttmbf && ttmb < 8) ttmb = -1;
2682 first_block = 0; 2682 first_block = 0;
2833 for(j = 0; j < 64; j++) s->block[i][j] += 128; 2833 for(j = 0; j < 64; j++) s->block[i][j] += 128;
2834 s->dsp.put_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2)); 2834 s->dsp.put_pixels_clamped(s->block[i], s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2));
2835 /* TODO: proper loop filtering */ 2835 /* TODO: proper loop filtering */
2836 if(v->pq >= 9 && v->overlap) { 2836 if(v->pq >= 9 && v->overlap) {
2837 if(v->a_avail) 2837 if(v->a_avail)
2838 vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2)); 2838 vc1_v_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2), (i<4) ? ((i&1)>>1) : (s->mb_y&1));
2839 if(v->c_avail) 2839 if(v->c_avail)
2840 vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2)); 2840 vc1_h_overlap(s->dest[dst_idx] + off, s->linesize >> ((i & 4) >> 2), (i<4) ? (i&1) : (s->mb_x&1));
2841 } 2841 }
2842 } else if(val) { 2842 } else if(val) {
2843 vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block); 2843 vc1_decode_p_block(v, s->block[i], i, mquant, ttmb, first_block);
2844 if(!v->ttmbf && ttmb < 8) ttmb = -1; 2844 if(!v->ttmbf && ttmb < 8) ttmb = -1;
2845 first_block = 0; 2845 first_block = 0;
2924 } 2924 }
2925 2925
2926 vc1_put_block(v, s->block); 2926 vc1_put_block(v, s->block);
2927 if(v->pq >= 9 && v->overlap) { /* XXX: do proper overlapping insted of loop filter */ 2927 if(v->pq >= 9 && v->overlap) { /* XXX: do proper overlapping insted of loop filter */
2928 if(!s->first_slice_line) { 2928 if(!s->first_slice_line) {
2929 vc1_v_overlap(s->dest[0], s->linesize); 2929 vc1_v_overlap(s->dest[0], s->linesize, 0);
2930 vc1_v_overlap(s->dest[0] + 8, s->linesize); 2930 vc1_v_overlap(s->dest[0] + 8, s->linesize, 0);
2931 vc1_v_overlap(s->dest[1], s->uvlinesize); 2931 vc1_v_overlap(s->dest[1], s->uvlinesize, s->mb_y&1);
2932 vc1_v_overlap(s->dest[2], s->uvlinesize); 2932 vc1_v_overlap(s->dest[2], s->uvlinesize, s->mb_y&1);
2933 } 2933 }
2934 vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize); 2934 vc1_v_overlap(s->dest[0] + 8 * s->linesize, s->linesize, 1);
2935 vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize); 2935 vc1_v_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize, 1);
2936 if(s->mb_x) { 2936 if(s->mb_x) {
2937 vc1_h_overlap(s->dest[0], s->linesize); 2937 vc1_h_overlap(s->dest[0], s->linesize, 0);
2938 vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize); 2938 vc1_h_overlap(s->dest[0] + 8 * s->linesize, s->linesize, 0);
2939 vc1_h_overlap(s->dest[1], s->uvlinesize); 2939 vc1_h_overlap(s->dest[1], s->uvlinesize, s->mb_x&1);
2940 vc1_h_overlap(s->dest[2], s->uvlinesize); 2940 vc1_h_overlap(s->dest[2], s->uvlinesize, s->mb_x&1);
2941 } 2941 }
2942 vc1_h_overlap(s->dest[0] + 8, s->linesize); 2942 vc1_h_overlap(s->dest[0] + 8, s->linesize, 1);
2943 vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize); 2943 vc1_h_overlap(s->dest[0] + 8 * s->linesize + 8, s->linesize, 1);
2944 } 2944 }
2945 2945
2946 if(get_bits_count(&s->gb) > v->bits) { 2946 if(get_bits_count(&s->gb) > v->bits) {
2947 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits); 2947 av_log(s->avctx, AV_LOG_ERROR, "Bits overconsumption: %i > %i\n", get_bits_count(&s->gb), v->bits);
2948 return; 2948 return;