3 * Copyright (c) 2000,2001 Fabrice Bellard
4 * Copyright (c) 2002-2004 Michael Niedermayer
6 * new motion estimation (X1/EPZS) by Michael Niedermayer <michaelni@gmx.at>
8 * This file is part of FFmpeg.
10 * FFmpeg is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU Lesser General Public
12 * License as published by the Free Software Foundation; either
13 * version 2.1 of the License, or (at your option) any later version.
15 * FFmpeg is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
20 * You should have received a copy of the GNU Lesser General Public
21 * License along with FFmpeg; if not, write to the Free Software
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
33 #include "libavutil/intmath.h"
37 #include "mpegvideo.h"
42 #define SQ(a) ((a)*(a))
46 #define P_TOPRIGHT P[3]
50 static inline int sad_hpel_motion_search(MpegEncContext * s,
51 int *mx_ptr, int *my_ptr, int dmin,
52 int src_index, int ref_index,
55 static inline int update_map_generation(MotionEstContext *c)
57 c->map_generation+= 1<<(ME_MAP_MV_BITS*2);
58 if(c->map_generation==0){
59 c->map_generation= 1<<(ME_MAP_MV_BITS*2);
60 memset(c->map, 0, sizeof(uint32_t)*ME_MAP_SIZE);
62 return c->map_generation;
65 /* shape adaptive search stuff */
66 typedef struct Minima{
72 static int minima_cmp(const void *a, const void *b){
73 const Minima *da = (const Minima *) a;
74 const Minima *db = (const Minima *) b;
76 return da->height - db->height;
79 #define FLAG_QPEL 1 //must be 1
83 static inline void init_ref(MotionEstContext *c, uint8_t *src[3], uint8_t *ref[3], uint8_t *ref2[3], int x, int y, int ref_index){
84 const int offset[3]= {
86 ((y*c->uvstride + x)>>1),
87 ((y*c->uvstride + x)>>1),
91 c->src[0][i]= src [i] + offset[i];
92 c->ref[0][i]= ref [i] + offset[i];
96 c->ref[ref_index][i]= ref2[i] + offset[i];
101 static int get_flags(MotionEstContext *c, int direct, int chroma){
102 return ((c->avctx->flags&CODEC_FLAG_QPEL) ? FLAG_QPEL : 0)
103 + (direct ? FLAG_DIRECT : 0)
104 + (chroma ? FLAG_CHROMA : 0);
107 static av_always_inline int cmp_direct_inline(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
108 const int size, const int h, int ref_index, int src_index,
109 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, int qpel){
110 MotionEstContext * const c= &s->me;
111 const int stride= c->stride;
112 const int hx= subx + (x<<(1+qpel));
113 const int hy= suby + (y<<(1+qpel));
114 uint8_t * const * const ref= c->ref[ref_index];
115 uint8_t * const * const src= c->src[src_index];
117 //FIXME check chroma 4mv, (no crashes ...)
118 assert(x >= c->xmin && hx <= c->xmax<<(qpel+1) && y >= c->ymin && hy <= c->ymax<<(qpel+1));
119 if(x >= c->xmin && hx <= c->xmax<<(qpel+1) && y >= c->ymin && hy <= c->ymax<<(qpel+1)){
120 const int time_pp= s->pp_time;
121 const int time_pb= s->pb_time;
122 const int mask= 2*qpel+1;
123 if(s->mv_type==MV_TYPE_8X8){
126 int fx = c->direct_basis_mv[i][0] + hx;
127 int fy = c->direct_basis_mv[i][1] + hy;
128 int bx = hx ? fx - c->co_located_mv[i][0] : c->co_located_mv[i][0]*(time_pb - time_pp)/time_pp + ((i &1)<<(qpel+4));
129 int by = hy ? fy - c->co_located_mv[i][1] : c->co_located_mv[i][1]*(time_pb - time_pp)/time_pp + ((i>>1)<<(qpel+4));
130 int fxy= (fx&mask) + ((fy&mask)<<(qpel+1));
131 int bxy= (bx&mask) + ((by&mask)<<(qpel+1));
133 uint8_t *dst= c->temp + 8*(i&1) + 8*stride*(i>>1);
135 c->qpel_put[1][fxy](dst, ref[0] + (fx>>2) + (fy>>2)*stride, stride);
136 c->qpel_avg[1][bxy](dst, ref[8] + (bx>>2) + (by>>2)*stride, stride);
138 c->hpel_put[1][fxy](dst, ref[0] + (fx>>1) + (fy>>1)*stride, stride, 8);
139 c->hpel_avg[1][bxy](dst, ref[8] + (bx>>1) + (by>>1)*stride, stride, 8);
143 int fx = c->direct_basis_mv[0][0] + hx;
144 int fy = c->direct_basis_mv[0][1] + hy;
145 int bx = hx ? fx - c->co_located_mv[0][0] : (c->co_located_mv[0][0]*(time_pb - time_pp)/time_pp);
146 int by = hy ? fy - c->co_located_mv[0][1] : (c->co_located_mv[0][1]*(time_pb - time_pp)/time_pp);
147 int fxy= (fx&mask) + ((fy&mask)<<(qpel+1));
148 int bxy= (bx&mask) + ((by&mask)<<(qpel+1));
151 c->qpel_put[1][fxy](c->temp , ref[0] + (fx>>2) + (fy>>2)*stride , stride);
152 c->qpel_put[1][fxy](c->temp + 8 , ref[0] + (fx>>2) + (fy>>2)*stride + 8 , stride);
153 c->qpel_put[1][fxy](c->temp + 8*stride, ref[0] + (fx>>2) + (fy>>2)*stride + 8*stride, stride);
154 c->qpel_put[1][fxy](c->temp + 8 + 8*stride, ref[0] + (fx>>2) + (fy>>2)*stride + 8 + 8*stride, stride);
155 c->qpel_avg[1][bxy](c->temp , ref[8] + (bx>>2) + (by>>2)*stride , stride);
156 c->qpel_avg[1][bxy](c->temp + 8 , ref[8] + (bx>>2) + (by>>2)*stride + 8 , stride);
157 c->qpel_avg[1][bxy](c->temp + 8*stride, ref[8] + (bx>>2) + (by>>2)*stride + 8*stride, stride);
158 c->qpel_avg[1][bxy](c->temp + 8 + 8*stride, ref[8] + (bx>>2) + (by>>2)*stride + 8 + 8*stride, stride);
160 assert((fx>>1) + 16*s->mb_x >= -16);
161 assert((fy>>1) + 16*s->mb_y >= -16);
162 assert((fx>>1) + 16*s->mb_x <= s->width);
163 assert((fy>>1) + 16*s->mb_y <= s->height);
164 assert((bx>>1) + 16*s->mb_x >= -16);
165 assert((by>>1) + 16*s->mb_y >= -16);
166 assert((bx>>1) + 16*s->mb_x <= s->width);
167 assert((by>>1) + 16*s->mb_y <= s->height);
169 c->hpel_put[0][fxy](c->temp, ref[0] + (fx>>1) + (fy>>1)*stride, stride, 16);
170 c->hpel_avg[0][bxy](c->temp, ref[8] + (bx>>1) + (by>>1)*stride, stride, 16);
173 d = cmp_func(s, c->temp, src[0], stride, 16);
179 static av_always_inline int cmp_inline(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
180 const int size, const int h, int ref_index, int src_index,
181 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, int qpel, int chroma){
182 MotionEstContext * const c= &s->me;
183 const int stride= c->stride;
184 const int uvstride= c->uvstride;
185 const int dxy= subx + (suby<<(1+qpel)); //FIXME log2_subpel?
186 const int hx= subx + (x<<(1+qpel));
187 const int hy= suby + (y<<(1+qpel));
188 uint8_t * const * const ref= c->ref[ref_index];
189 uint8_t * const * const src= c->src[src_index];
191 //FIXME check chroma 4mv, (no crashes ...)
192 int uvdxy; /* no, it might not be used uninitialized */
195 c->qpel_put[size][dxy](c->temp, ref[0] + x + y*stride, stride); //FIXME prototype (add h)
201 uvdxy= (cx&1) + 2*(cy&1);
202 //FIXME x/y wrong, but mpeg4 qpel is sick anyway, we should drop as much of it as possible in favor for h264
205 c->hpel_put[size][dxy](c->temp, ref[0] + x + y*stride, stride, h);
207 uvdxy= dxy | (x&1) | (2*(y&1));
209 d = cmp_func(s, c->temp, src[0], stride, h);
211 d = cmp_func(s, src[0], ref[0] + x + y*stride, stride, h);
213 uvdxy= (x&1) + 2*(y&1);
216 uint8_t * const uvtemp= c->temp + 16*stride;
217 c->hpel_put[size+1][uvdxy](uvtemp , ref[1] + (x>>1) + (y>>1)*uvstride, uvstride, h>>1);
218 c->hpel_put[size+1][uvdxy](uvtemp+8, ref[2] + (x>>1) + (y>>1)*uvstride, uvstride, h>>1);
219 d += chroma_cmp_func(s, uvtemp , src[1], uvstride, h>>1);
220 d += chroma_cmp_func(s, uvtemp+8, src[2], uvstride, h>>1);
225 static int cmp_simple(MpegEncContext *s, const int x, const int y,
226 int ref_index, int src_index,
227 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func){
228 return cmp_inline(s,x,y,0,0,0,16,ref_index,src_index, cmp_func, chroma_cmp_func, 0, 0);
231 static int cmp_fpel_internal(MpegEncContext *s, const int x, const int y,
232 const int size, const int h, int ref_index, int src_index,
233 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
234 if(flags&FLAG_DIRECT){
235 return cmp_direct_inline(s,x,y,0,0,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, flags&FLAG_QPEL);
237 return cmp_inline(s,x,y,0,0,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 0, flags&FLAG_CHROMA);
241 static int cmp_internal(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
242 const int size, const int h, int ref_index, int src_index,
243 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
244 if(flags&FLAG_DIRECT){
245 return cmp_direct_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, flags&FLAG_QPEL);
247 return cmp_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, flags&FLAG_QPEL, flags&FLAG_CHROMA);
251 /*! \brief compares a block (either a full macroblock or a partition thereof)
252 against a proposed motion-compensated prediction of that block
254 static av_always_inline int cmp(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
255 const int size, const int h, int ref_index, int src_index,
256 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
257 if(av_builtin_constant_p(flags) && av_builtin_constant_p(h) && av_builtin_constant_p(size)
258 && av_builtin_constant_p(subx) && av_builtin_constant_p(suby)
259 && flags==0 && h==16 && size==0 && subx==0 && suby==0){
260 return cmp_simple(s,x,y,ref_index,src_index, cmp_func, chroma_cmp_func);
261 }else if(av_builtin_constant_p(subx) && av_builtin_constant_p(suby)
262 && subx==0 && suby==0){
263 return cmp_fpel_internal(s,x,y,size,h,ref_index,src_index, cmp_func, chroma_cmp_func,flags);
265 return cmp_internal(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, flags);
269 static int cmp_hpel(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
270 const int size, const int h, int ref_index, int src_index,
271 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
272 if(flags&FLAG_DIRECT){
273 return cmp_direct_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 0);
275 return cmp_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 0, flags&FLAG_CHROMA);
279 static int cmp_qpel(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
280 const int size, const int h, int ref_index, int src_index,
281 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
282 if(flags&FLAG_DIRECT){
283 return cmp_direct_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 1);
285 return cmp_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 1, flags&FLAG_CHROMA);
289 #include "motion_est_template.c"
291 static int zero_cmp(void *s, uint8_t *a, uint8_t *b, int stride, int h){
295 static void zero_hpel(uint8_t *a, const uint8_t *b, int stride, int h){
298 int ff_init_me(MpegEncContext *s){
299 MotionEstContext * const c= &s->me;
300 int cache_size= FFMIN(ME_MAP_SIZE>>ME_MAP_SHIFT, 1<<ME_MAP_SHIFT);
301 int dia_size= FFMAX(FFABS(s->avctx->dia_size)&255, FFABS(s->avctx->pre_dia_size)&255);
303 if(FFMIN(s->avctx->dia_size, s->avctx->pre_dia_size) < -ME_MAP_SIZE){
304 av_log(s->avctx, AV_LOG_ERROR, "ME_MAP size is too small for SAB diamond\n");
307 //special case of snow is needed because snow uses its own iterative ME code
308 if(s->me_method!=ME_ZERO && s->me_method!=ME_EPZS && s->me_method!=ME_X1 && s->avctx->codec_id != CODEC_ID_SNOW){
309 av_log(s->avctx, AV_LOG_ERROR, "me_method is only allowed to be set to zero and epzs; for hex,umh,full and others see dia_size\n");
315 if(cache_size < 2*dia_size && !c->stride){
316 av_log(s->avctx, AV_LOG_INFO, "ME_MAP size may be a little small for the selected diamond size\n");
319 ff_set_cmp(&s->dsp, s->dsp.me_pre_cmp, c->avctx->me_pre_cmp);
320 ff_set_cmp(&s->dsp, s->dsp.me_cmp, c->avctx->me_cmp);
321 ff_set_cmp(&s->dsp, s->dsp.me_sub_cmp, c->avctx->me_sub_cmp);
322 ff_set_cmp(&s->dsp, s->dsp.mb_cmp, c->avctx->mb_cmp);
324 c->flags = get_flags(c, 0, c->avctx->me_cmp &FF_CMP_CHROMA);
325 c->sub_flags= get_flags(c, 0, c->avctx->me_sub_cmp&FF_CMP_CHROMA);
326 c->mb_flags = get_flags(c, 0, c->avctx->mb_cmp &FF_CMP_CHROMA);
328 /*FIXME s->no_rounding b_type*/
329 if(s->flags&CODEC_FLAG_QPEL){
330 c->sub_motion_search= qpel_motion_search;
331 c->qpel_avg= s->dsp.avg_qpel_pixels_tab;
332 if(s->no_rounding) c->qpel_put= s->dsp.put_no_rnd_qpel_pixels_tab;
333 else c->qpel_put= s->dsp.put_qpel_pixels_tab;
335 if(c->avctx->me_sub_cmp&FF_CMP_CHROMA)
336 c->sub_motion_search= hpel_motion_search;
337 else if( c->avctx->me_sub_cmp == FF_CMP_SAD
338 && c->avctx-> me_cmp == FF_CMP_SAD
339 && c->avctx-> mb_cmp == FF_CMP_SAD)
340 c->sub_motion_search= sad_hpel_motion_search; // 2050 vs. 2450 cycles
342 c->sub_motion_search= hpel_motion_search;
344 c->hpel_avg= s->dsp.avg_pixels_tab;
345 if(s->no_rounding) c->hpel_put= s->dsp.put_no_rnd_pixels_tab;
346 else c->hpel_put= s->dsp.put_pixels_tab;
349 c->stride = s->linesize;
350 c->uvstride= s->uvlinesize;
352 c->stride = 16*s->mb_width + 32;
353 c->uvstride= 8*s->mb_width + 16;
356 /* 8x8 fullpel search would need a 4x4 chroma compare, which we do
357 * not have yet, and even if we had, the motion estimation code
358 * does not expect it. */
359 if(s->codec_id != CODEC_ID_SNOW){
360 if((c->avctx->me_cmp&FF_CMP_CHROMA)/* && !s->dsp.me_cmp[2]*/){
361 s->dsp.me_cmp[2]= zero_cmp;
363 if((c->avctx->me_sub_cmp&FF_CMP_CHROMA) && !s->dsp.me_sub_cmp[2]){
364 s->dsp.me_sub_cmp[2]= zero_cmp;
366 c->hpel_put[2][0]= c->hpel_put[2][1]=
367 c->hpel_put[2][2]= c->hpel_put[2][3]= zero_hpel;
370 if(s->codec_id == CODEC_ID_H261){
371 c->sub_motion_search= no_sub_motion_search;
378 static int pix_dev(uint8_t * pix, int line_size, int mean)
383 for (i = 0; i < 16; i++) {
384 for (j = 0; j < 16; j += 8) {
385 s += FFABS(pix[0]-mean);
386 s += FFABS(pix[1]-mean);
387 s += FFABS(pix[2]-mean);
388 s += FFABS(pix[3]-mean);
389 s += FFABS(pix[4]-mean);
390 s += FFABS(pix[5]-mean);
391 s += FFABS(pix[6]-mean);
392 s += FFABS(pix[7]-mean);
395 pix += line_size - 16;
401 static inline void no_motion_search(MpegEncContext * s,
402 int *mx_ptr, int *my_ptr)
404 *mx_ptr = 16 * s->mb_x;
405 *my_ptr = 16 * s->mb_y;
408 #define Z_THRESHOLD 256
410 #define CHECK_SAD_HALF_MV(suffix, x, y) \
412 d= s->dsp.pix_abs[size][(x?1:0)+(y?2:0)](NULL, pix, ptr+((x)>>1), stride, h);\
413 d += (mv_penalty[pen_x + x] + mv_penalty[pen_y + y])*penalty_factor;\
414 COPY3_IF_LT(dminh, d, dx, x, dy, y)\
417 static inline int sad_hpel_motion_search(MpegEncContext * s,
418 int *mx_ptr, int *my_ptr, int dmin,
419 int src_index, int ref_index,
422 MotionEstContext * const c= &s->me;
423 const int penalty_factor= c->sub_penalty_factor;
426 int stride= c->stride;
427 const int flags= c->sub_flags;
440 pix = c->src[src_index][0];
444 ptr = c->ref[ref_index][0] + (my * stride) + mx;
448 if (mx > xmin && mx < xmax &&
449 my > ymin && my < ymax) {
452 const int index= (my<<ME_MAP_SHIFT) + mx;
453 const int t= score_map[(index-(1<<ME_MAP_SHIFT))&(ME_MAP_SIZE-1)];
454 const int l= score_map[(index- 1 )&(ME_MAP_SIZE-1)];
455 const int r= score_map[(index+ 1 )&(ME_MAP_SIZE-1)];
456 const int b= score_map[(index+(1<<ME_MAP_SHIFT))&(ME_MAP_SIZE-1)];
466 CHECK_SAD_HALF_MV(y2 , 0, -1)
468 CHECK_SAD_HALF_MV(xy2, -1, -1)
470 CHECK_SAD_HALF_MV(xy2, +1, -1)
474 CHECK_SAD_HALF_MV(xy2, -1, +1)
476 CHECK_SAD_HALF_MV(x2 , -1, 0)
478 CHECK_SAD_HALF_MV(xy2, +1, -1)
480 CHECK_SAD_HALF_MV(xy2, -1, -1)
484 CHECK_SAD_HALF_MV(xy2, +1, +1)
486 CHECK_SAD_HALF_MV(x2 , +1, 0)
491 CHECK_SAD_HALF_MV(xy2, -1, -1)
495 CHECK_SAD_HALF_MV(xy2, +1, +1)
497 CHECK_SAD_HALF_MV(x2 , -1, 0)
498 CHECK_SAD_HALF_MV(xy2, -1, +1)
501 CHECK_SAD_HALF_MV(xy2, +1, -1)
505 CHECK_SAD_HALF_MV(xy2, -1, +1)
507 CHECK_SAD_HALF_MV(x2 , +1, 0)
508 CHECK_SAD_HALF_MV(xy2, +1, +1)
510 CHECK_SAD_HALF_MV(y2 , 0, +1)
525 static inline void set_p_mv_tables(MpegEncContext * s, int mx, int my, int mv4)
527 const int xy= s->mb_x + s->mb_y*s->mb_stride;
529 s->p_mv_table[xy][0] = mx;
530 s->p_mv_table[xy][1] = my;
532 /* has already been set to the 4 MV if 4MV is done */
534 int mot_xy= s->block_index[0];
536 s->current_picture.motion_val[0][mot_xy ][0]= mx;
537 s->current_picture.motion_val[0][mot_xy ][1]= my;
538 s->current_picture.motion_val[0][mot_xy+1][0]= mx;
539 s->current_picture.motion_val[0][mot_xy+1][1]= my;
541 mot_xy += s->b8_stride;
542 s->current_picture.motion_val[0][mot_xy ][0]= mx;
543 s->current_picture.motion_val[0][mot_xy ][1]= my;
544 s->current_picture.motion_val[0][mot_xy+1][0]= mx;
545 s->current_picture.motion_val[0][mot_xy+1][1]= my;
550 * get fullpel ME search limits.
552 static inline void get_limits(MpegEncContext *s, int x, int y)
554 MotionEstContext * const c= &s->me;
555 int range= c->avctx->me_range >> (1 + !!(c->flags&FLAG_QPEL));
557 if(c->avctx->me_range) c->range= c->avctx->me_range >> 1;
560 if (s->unrestricted_mv) {
563 c->xmax = - x + s->mb_width *16;
564 c->ymax = - y + s->mb_height*16;
565 } else if (s->out_format == FMT_H261){
566 // Search range of H261 is different from other codec standards
567 c->xmin = (x > 15) ? - 15 : 0;
568 c->ymin = (y > 15) ? - 15 : 0;
569 c->xmax = (x < s->mb_width * 16 - 16) ? 15 : 0;
570 c->ymax = (y < s->mb_height * 16 - 16) ? 15 : 0;
574 c->xmax = - x + s->mb_width *16 - 16;
575 c->ymax = - y + s->mb_height*16 - 16;
578 c->xmin = FFMAX(c->xmin,-range);
579 c->xmax = FFMIN(c->xmax, range);
580 c->ymin = FFMAX(c->ymin,-range);
581 c->ymax = FFMIN(c->ymax, range);
585 static inline void init_mv4_ref(MotionEstContext *c){
586 const int stride= c->stride;
588 c->ref[1][0] = c->ref[0][0] + 8;
589 c->ref[2][0] = c->ref[0][0] + 8*stride;
590 c->ref[3][0] = c->ref[2][0] + 8;
591 c->src[1][0] = c->src[0][0] + 8;
592 c->src[2][0] = c->src[0][0] + 8*stride;
593 c->src[3][0] = c->src[2][0] + 8;
596 static inline int h263_mv4_search(MpegEncContext *s, int mx, int my, int shift)
598 MotionEstContext * const c= &s->me;
603 int dmin_sum=0, mx4_sum=0, my4_sum=0;
605 const int stride= c->stride;
606 uint8_t *mv_penalty= c->current_mv_penalty;
610 for(block=0; block<4; block++){
612 int pred_x4, pred_y4;
614 static const int off[4]= {2, 1, 1, -1};
615 const int mot_stride = s->b8_stride;
616 const int mot_xy = s->block_index[block];
618 P_LEFT[0] = s->current_picture.motion_val[0][mot_xy - 1][0];
619 P_LEFT[1] = s->current_picture.motion_val[0][mot_xy - 1][1];
621 if(P_LEFT[0] > (c->xmax<<shift)) P_LEFT[0] = (c->xmax<<shift);
623 /* special case for first line */
624 if (s->first_slice_line && block<2) {
625 c->pred_x= pred_x4= P_LEFT[0];
626 c->pred_y= pred_y4= P_LEFT[1];
628 P_TOP[0] = s->current_picture.motion_val[0][mot_xy - mot_stride ][0];
629 P_TOP[1] = s->current_picture.motion_val[0][mot_xy - mot_stride ][1];
630 P_TOPRIGHT[0] = s->current_picture.motion_val[0][mot_xy - mot_stride + off[block]][0];
631 P_TOPRIGHT[1] = s->current_picture.motion_val[0][mot_xy - mot_stride + off[block]][1];
632 if(P_TOP[1] > (c->ymax<<shift)) P_TOP[1] = (c->ymax<<shift);
633 if(P_TOPRIGHT[0] < (c->xmin<<shift)) P_TOPRIGHT[0]= (c->xmin<<shift);
634 if(P_TOPRIGHT[0] > (c->xmax<<shift)) P_TOPRIGHT[0]= (c->xmax<<shift);
635 if(P_TOPRIGHT[1] > (c->ymax<<shift)) P_TOPRIGHT[1]= (c->ymax<<shift);
637 P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
638 P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
640 c->pred_x= pred_x4 = P_MEDIAN[0];
641 c->pred_y= pred_y4 = P_MEDIAN[1];
646 dmin4 = epzs_motion_search4(s, &mx4, &my4, P, block, block, s->p_mv_table, (1<<16)>>shift);
648 dmin4= c->sub_motion_search(s, &mx4, &my4, dmin4, block, block, size, h);
650 if(s->dsp.me_sub_cmp[0] != s->dsp.mb_cmp[0]){
652 const int offset= ((block&1) + (block>>1)*stride)*8;
653 uint8_t *dest_y = c->scratchpad + offset;
654 if(s->quarter_sample){
655 uint8_t *ref= c->ref[block][0] + (mx4>>2) + (my4>>2)*stride;
656 dxy = ((my4 & 3) << 2) | (mx4 & 3);
659 s->dsp.put_no_rnd_qpel_pixels_tab[1][dxy](dest_y , ref , stride);
661 s->dsp.put_qpel_pixels_tab [1][dxy](dest_y , ref , stride);
663 uint8_t *ref= c->ref[block][0] + (mx4>>1) + (my4>>1)*stride;
664 dxy = ((my4 & 1) << 1) | (mx4 & 1);
667 s->dsp.put_no_rnd_pixels_tab[1][dxy](dest_y , ref , stride, h);
669 s->dsp.put_pixels_tab [1][dxy](dest_y , ref , stride, h);
671 dmin_sum+= (mv_penalty[mx4-pred_x4] + mv_penalty[my4-pred_y4])*c->mb_penalty_factor;
675 if(s->quarter_sample){
683 s->current_picture.motion_val[0][ s->block_index[block] ][0]= mx4;
684 s->current_picture.motion_val[0][ s->block_index[block] ][1]= my4;
686 if(mx4 != mx || my4 != my) same=0;
692 if(s->dsp.me_sub_cmp[0] != s->dsp.mb_cmp[0]){
693 dmin_sum += s->dsp.mb_cmp[0](s, s->new_picture.data[0] + s->mb_x*16 + s->mb_y*16*stride, c->scratchpad, stride, 16);
696 if(c->avctx->mb_cmp&FF_CMP_CHROMA){
701 mx= ff_h263_round_chroma(mx4_sum);
702 my= ff_h263_round_chroma(my4_sum);
703 dxy = ((my & 1) << 1) | (mx & 1);
705 offset= (s->mb_x*8 + (mx>>1)) + (s->mb_y*8 + (my>>1))*s->uvlinesize;
708 s->dsp.put_no_rnd_pixels_tab[1][dxy](c->scratchpad , s->last_picture.data[1] + offset, s->uvlinesize, 8);
709 s->dsp.put_no_rnd_pixels_tab[1][dxy](c->scratchpad+8 , s->last_picture.data[2] + offset, s->uvlinesize, 8);
711 s->dsp.put_pixels_tab [1][dxy](c->scratchpad , s->last_picture.data[1] + offset, s->uvlinesize, 8);
712 s->dsp.put_pixels_tab [1][dxy](c->scratchpad+8 , s->last_picture.data[2] + offset, s->uvlinesize, 8);
715 dmin_sum += s->dsp.mb_cmp[1](s, s->new_picture.data[1] + s->mb_x*8 + s->mb_y*8*s->uvlinesize, c->scratchpad , s->uvlinesize, 8);
716 dmin_sum += s->dsp.mb_cmp[1](s, s->new_picture.data[2] + s->mb_x*8 + s->mb_y*8*s->uvlinesize, c->scratchpad+8, s->uvlinesize, 8);
722 switch(c->avctx->mb_cmp&0xFF){
724 return dmin_sum+ 32*s->qscale*s->qscale;*/
728 return dmin_sum+ 11*c->mb_penalty_factor;
732 static inline void init_interlaced_ref(MpegEncContext *s, int ref_index){
733 MotionEstContext * const c= &s->me;
735 c->ref[1+ref_index][0] = c->ref[0+ref_index][0] + s->linesize;
736 c->src[1][0] = c->src[0][0] + s->linesize;
737 if(c->flags & FLAG_CHROMA){
738 c->ref[1+ref_index][1] = c->ref[0+ref_index][1] + s->uvlinesize;
739 c->ref[1+ref_index][2] = c->ref[0+ref_index][2] + s->uvlinesize;
740 c->src[1][1] = c->src[0][1] + s->uvlinesize;
741 c->src[1][2] = c->src[0][2] + s->uvlinesize;
745 static int interlaced_search(MpegEncContext *s, int ref_index,
746 int16_t (*mv_tables[2][2])[2], uint8_t *field_select_tables[2], int mx, int my, int user_field_select)
748 MotionEstContext * const c= &s->me;
753 uint8_t * const mv_penalty= c->current_mv_penalty;
755 const int stride= 2*s->linesize;
757 const int mot_stride= s->mb_stride;
758 const int xy= s->mb_x + s->mb_y*mot_stride;
764 init_interlaced_ref(s, ref_index);
766 for(block=0; block<2; block++){
768 int best_dmin= INT_MAX;
771 for(field_select=0; field_select<2; field_select++){
772 int dmin, mx_i, my_i;
773 int16_t (*mv_table)[2]= mv_tables[block][field_select];
775 if(user_field_select){
776 assert(field_select==0 || field_select==1);
777 assert(field_select_tables[block][xy]==0 || field_select_tables[block][xy]==1);
778 if(field_select_tables[block][xy] != field_select)
782 P_LEFT[0] = mv_table[xy - 1][0];
783 P_LEFT[1] = mv_table[xy - 1][1];
784 if(P_LEFT[0] > (c->xmax<<1)) P_LEFT[0] = (c->xmax<<1);
786 c->pred_x= P_LEFT[0];
787 c->pred_y= P_LEFT[1];
789 if(!s->first_slice_line){
790 P_TOP[0] = mv_table[xy - mot_stride][0];
791 P_TOP[1] = mv_table[xy - mot_stride][1];
792 P_TOPRIGHT[0] = mv_table[xy - mot_stride + 1][0];
793 P_TOPRIGHT[1] = mv_table[xy - mot_stride + 1][1];
794 if(P_TOP[1] > (c->ymax<<1)) P_TOP[1] = (c->ymax<<1);
795 if(P_TOPRIGHT[0] < (c->xmin<<1)) P_TOPRIGHT[0]= (c->xmin<<1);
796 if(P_TOPRIGHT[0] > (c->xmax<<1)) P_TOPRIGHT[0]= (c->xmax<<1);
797 if(P_TOPRIGHT[1] > (c->ymax<<1)) P_TOPRIGHT[1]= (c->ymax<<1);
799 P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
800 P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
802 P_MV1[0]= mx; //FIXME not correct if block != field_select
805 dmin = epzs_motion_search2(s, &mx_i, &my_i, P, block, field_select+ref_index, mv_table, (1<<16)>>1);
807 dmin= c->sub_motion_search(s, &mx_i, &my_i, dmin, block, field_select+ref_index, size, h);
809 mv_table[xy][0]= mx_i;
810 mv_table[xy][1]= my_i;
812 if(s->dsp.me_sub_cmp[0] != s->dsp.mb_cmp[0]){
816 uint8_t *ref= c->ref[field_select+ref_index][0] + (mx_i>>1) + (my_i>>1)*stride;
817 dxy = ((my_i & 1) << 1) | (mx_i & 1);
820 s->dsp.put_no_rnd_pixels_tab[size][dxy](c->scratchpad, ref , stride, h);
822 s->dsp.put_pixels_tab [size][dxy](c->scratchpad, ref , stride, h);
824 dmin= s->dsp.mb_cmp[size](s, c->src[block][0], c->scratchpad, stride, h);
825 dmin+= (mv_penalty[mx_i-c->pred_x] + mv_penalty[my_i-c->pred_y] + 1)*c->mb_penalty_factor;
827 dmin+= c->mb_penalty_factor; //field_select bits
829 dmin += field_select != block; //slightly prefer same field
831 if(dmin < best_dmin){
833 best_field= field_select;
837 int16_t (*mv_table)[2]= mv_tables[block][best_field];
839 if(mv_table[xy][0] != mx) same=0; //FIXME check if these checks work and are any good at all
840 if(mv_table[xy][1]&1) same=0;
841 if(mv_table[xy][1]*2 != my) same=0;
842 if(best_field != block) same=0;
845 field_select_tables[block][xy]= best_field;
846 dmin_sum += best_dmin;
857 switch(c->avctx->mb_cmp&0xFF){
859 return dmin_sum+ 32*s->qscale*s->qscale;*/
863 return dmin_sum+ 11*c->mb_penalty_factor;
867 static void clip_input_mv(MpegEncContext * s, int16_t *mv, int interlaced){
868 int ymax= s->me.ymax>>interlaced;
869 int ymin= s->me.ymin>>interlaced;
871 if(mv[0] < s->me.xmin) mv[0] = s->me.xmin;
872 if(mv[0] > s->me.xmax) mv[0] = s->me.xmax;
873 if(mv[1] < ymin) mv[1] = ymin;
874 if(mv[1] > ymax) mv[1] = ymax;
877 static inline int check_input_motion(MpegEncContext * s, int mb_x, int mb_y, int p_type){
878 MotionEstContext * const c= &s->me;
879 Picture *p= s->current_picture_ptr;
880 int mb_xy= mb_x + mb_y*s->mb_stride;
881 int xy= 2*mb_x + 2*mb_y*s->b8_stride;
882 int mb_type= s->current_picture.mb_type[mb_xy];
884 int shift= (flags&FLAG_QPEL) + 1;
885 int mask= (1<<shift)-1;
888 me_cmp_func cmpf= s->dsp.sse[0];
889 me_cmp_func chroma_cmpf= s->dsp.sse[1];
891 if(p_type && USES_LIST(mb_type, 1)){
892 av_log(c->avctx, AV_LOG_ERROR, "backward motion vector in P frame\n");
895 assert(IS_INTRA(mb_type) || USES_LIST(mb_type,0) || USES_LIST(mb_type,1));
898 int xy= s->block_index[i];
899 clip_input_mv(s, p->motion_val[0][xy], !!IS_INTERLACED(mb_type));
900 clip_input_mv(s, p->motion_val[1][xy], !!IS_INTERLACED(mb_type));
903 if(IS_INTERLACED(mb_type)){
904 int xy2= xy + s->b8_stride;
905 s->mb_type[mb_xy]=CANDIDATE_MB_TYPE_INTRA;
909 if(!(s->flags & CODEC_FLAG_INTERLACED_ME)){
910 av_log(c->avctx, AV_LOG_ERROR, "Interlaced macroblock selected but interlaced motion estimation disabled\n");
914 if(USES_LIST(mb_type, 0)){
915 int field_select0= p->ref_index[0][4*mb_xy ];
916 int field_select1= p->ref_index[0][4*mb_xy+2];
917 assert(field_select0==0 ||field_select0==1);
918 assert(field_select1==0 ||field_select1==1);
919 init_interlaced_ref(s, 0);
922 s->p_field_select_table[0][mb_xy]= field_select0;
923 s->p_field_select_table[1][mb_xy]= field_select1;
924 *(uint32_t*)s->p_field_mv_table[0][field_select0][mb_xy]= *(uint32_t*)p->motion_val[0][xy ];
925 *(uint32_t*)s->p_field_mv_table[1][field_select1][mb_xy]= *(uint32_t*)p->motion_val[0][xy2];
926 s->mb_type[mb_xy]=CANDIDATE_MB_TYPE_INTER_I;
928 s->b_field_select_table[0][0][mb_xy]= field_select0;
929 s->b_field_select_table[0][1][mb_xy]= field_select1;
930 *(uint32_t*)s->b_field_mv_table[0][0][field_select0][mb_xy]= *(uint32_t*)p->motion_val[0][xy ];
931 *(uint32_t*)s->b_field_mv_table[0][1][field_select1][mb_xy]= *(uint32_t*)p->motion_val[0][xy2];
932 s->mb_type[mb_xy]= CANDIDATE_MB_TYPE_FORWARD_I;
935 x= p->motion_val[0][xy ][0];
936 y= p->motion_val[0][xy ][1];
937 d = cmp(s, x>>shift, y>>shift, x&mask, y&mask, 0, 8, field_select0, 0, cmpf, chroma_cmpf, flags);
938 x= p->motion_val[0][xy2][0];
939 y= p->motion_val[0][xy2][1];
940 d+= cmp(s, x>>shift, y>>shift, x&mask, y&mask, 0, 8, field_select1, 1, cmpf, chroma_cmpf, flags);
942 if(USES_LIST(mb_type, 1)){
943 int field_select0= p->ref_index[1][4*mb_xy ];
944 int field_select1= p->ref_index[1][4*mb_xy+2];
945 assert(field_select0==0 ||field_select0==1);
946 assert(field_select1==0 ||field_select1==1);
947 init_interlaced_ref(s, 2);
949 s->b_field_select_table[1][0][mb_xy]= field_select0;
950 s->b_field_select_table[1][1][mb_xy]= field_select1;
951 *(uint32_t*)s->b_field_mv_table[1][0][field_select0][mb_xy]= *(uint32_t*)p->motion_val[1][xy ];
952 *(uint32_t*)s->b_field_mv_table[1][1][field_select1][mb_xy]= *(uint32_t*)p->motion_val[1][xy2];
953 if(USES_LIST(mb_type, 0)){
954 s->mb_type[mb_xy]= CANDIDATE_MB_TYPE_BIDIR_I;
956 s->mb_type[mb_xy]= CANDIDATE_MB_TYPE_BACKWARD_I;
959 x= p->motion_val[1][xy ][0];
960 y= p->motion_val[1][xy ][1];
961 d = cmp(s, x>>shift, y>>shift, x&mask, y&mask, 0, 8, field_select0+2, 0, cmpf, chroma_cmpf, flags);
962 x= p->motion_val[1][xy2][0];
963 y= p->motion_val[1][xy2][1];
964 d+= cmp(s, x>>shift, y>>shift, x&mask, y&mask, 0, 8, field_select1+2, 1, cmpf, chroma_cmpf, flags);
969 }else if(IS_8X8(mb_type)){
970 if(!(s->flags & CODEC_FLAG_4MV)){
971 av_log(c->avctx, AV_LOG_ERROR, "4MV macroblock selected but 4MV encoding disabled\n");
975 chroma_cmpf= s->dsp.sse[1];
978 xy= s->block_index[i];
979 x= p->motion_val[0][xy][0];
980 y= p->motion_val[0][xy][1];
981 d+= cmp(s, x>>shift, y>>shift, x&mask, y&mask, 1, 8, i, i, cmpf, chroma_cmpf, flags);
983 s->mb_type[mb_xy]=CANDIDATE_MB_TYPE_INTER4V;
985 if(USES_LIST(mb_type, 0)){
987 *(uint32_t*)s->p_mv_table[mb_xy]= *(uint32_t*)p->motion_val[0][xy];
988 s->mb_type[mb_xy]=CANDIDATE_MB_TYPE_INTER;
989 }else if(USES_LIST(mb_type, 1)){
990 *(uint32_t*)s->b_bidir_forw_mv_table[mb_xy]= *(uint32_t*)p->motion_val[0][xy];
991 *(uint32_t*)s->b_bidir_back_mv_table[mb_xy]= *(uint32_t*)p->motion_val[1][xy];
992 s->mb_type[mb_xy]=CANDIDATE_MB_TYPE_BIDIR;
994 *(uint32_t*)s->b_forw_mv_table[mb_xy]= *(uint32_t*)p->motion_val[0][xy];
995 s->mb_type[mb_xy]=CANDIDATE_MB_TYPE_FORWARD;
997 x= p->motion_val[0][xy][0];
998 y= p->motion_val[0][xy][1];
999 d = cmp(s, x>>shift, y>>shift, x&mask, y&mask, 0, 16, 0, 0, cmpf, chroma_cmpf, flags);
1000 }else if(USES_LIST(mb_type, 1)){
1001 *(uint32_t*)s->b_back_mv_table[mb_xy]= *(uint32_t*)p->motion_val[1][xy];
1002 s->mb_type[mb_xy]=CANDIDATE_MB_TYPE_BACKWARD;
1004 x= p->motion_val[1][xy][0];
1005 y= p->motion_val[1][xy][1];
1006 d = cmp(s, x>>shift, y>>shift, x&mask, y&mask, 0, 16, 2, 0, cmpf, chroma_cmpf, flags);
1008 s->mb_type[mb_xy]=CANDIDATE_MB_TYPE_INTRA;
1013 void ff_estimate_p_frame_motion(MpegEncContext * s,
1016 MotionEstContext * const c= &s->me;
1017 uint8_t *pix, *ppix;
1018 int sum, mx, my, dmin;
1019 int varc; ///< the variance of the block (sum of squared (p[y][x]-average))
1020 int vard; ///< sum of squared differences with the estimated motion vector
1022 const int shift= 1+s->quarter_sample;
1024 Picture * const pic= &s->current_picture;
1026 init_ref(c, s->new_picture.data, s->last_picture.data, NULL, 16*mb_x, 16*mb_y, 0);
1028 assert(s->quarter_sample==0 || s->quarter_sample==1);
1029 assert(s->linesize == c->stride);
1030 assert(s->uvlinesize == c->uvstride);
1032 c->penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_cmp);
1033 c->sub_penalty_factor= get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_sub_cmp);
1034 c->mb_penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->mb_cmp);
1035 c->current_mv_penalty= c->mv_penalty[s->f_code] + MAX_MV;
1037 get_limits(s, 16*mb_x, 16*mb_y);
1040 /* intra / predictive decision */
1042 sum = s->dsp.pix_sum(pix, s->linesize);
1043 varc = s->dsp.pix_norm1(pix, s->linesize) - (((unsigned)(sum*sum))>>8) + 500;
1045 pic->mb_mean[s->mb_stride * mb_y + mb_x] = (sum+128)>>8;
1046 pic->mb_var [s->mb_stride * mb_y + mb_x] = (varc+128)>>8;
1047 c->mb_var_sum_temp += (varc+128)>>8;
1049 if(c->avctx->me_threshold){
1050 vard= check_input_motion(s, mb_x, mb_y, 1);
1052 if((vard+128)>>8 < c->avctx->me_threshold){
1053 int p_score= FFMIN(vard, varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*100);
1054 int i_score= varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*20;
1055 pic->mc_mb_var[s->mb_stride * mb_y + mb_x] = (vard+128)>>8;
1056 c->mc_mb_var_sum_temp += (vard+128)>>8;
1057 c->scene_change_score+= ff_sqrt(p_score) - ff_sqrt(i_score);
1060 if((vard+128)>>8 < c->avctx->mb_threshold)
1061 mb_type= s->mb_type[mb_x + mb_y*s->mb_stride];
1064 switch(s->me_method) {
1067 no_motion_search(s, &mx, &my);
1075 const int mot_stride = s->b8_stride;
1076 const int mot_xy = s->block_index[0];
1078 P_LEFT[0] = s->current_picture.motion_val[0][mot_xy - 1][0];
1079 P_LEFT[1] = s->current_picture.motion_val[0][mot_xy - 1][1];
1081 if(P_LEFT[0] > (c->xmax<<shift)) P_LEFT[0] = (c->xmax<<shift);
1083 if(!s->first_slice_line) {
1084 P_TOP[0] = s->current_picture.motion_val[0][mot_xy - mot_stride ][0];
1085 P_TOP[1] = s->current_picture.motion_val[0][mot_xy - mot_stride ][1];
1086 P_TOPRIGHT[0] = s->current_picture.motion_val[0][mot_xy - mot_stride + 2][0];
1087 P_TOPRIGHT[1] = s->current_picture.motion_val[0][mot_xy - mot_stride + 2][1];
1088 if(P_TOP[1] > (c->ymax<<shift)) P_TOP[1] = (c->ymax<<shift);
1089 if(P_TOPRIGHT[0] < (c->xmin<<shift)) P_TOPRIGHT[0]= (c->xmin<<shift);
1090 if(P_TOPRIGHT[1] > (c->ymax<<shift)) P_TOPRIGHT[1]= (c->ymax<<shift);
1092 P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
1093 P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
1095 if(s->out_format == FMT_H263){
1096 c->pred_x = P_MEDIAN[0];
1097 c->pred_y = P_MEDIAN[1];
1098 }else { /* mpeg1 at least */
1099 c->pred_x= P_LEFT[0];
1100 c->pred_y= P_LEFT[1];
1103 c->pred_x= P_LEFT[0];
1104 c->pred_y= P_LEFT[1];
1108 dmin = ff_epzs_motion_search(s, &mx, &my, P, 0, 0, s->p_mv_table, (1<<16)>>shift, 0, 16);
1113 /* At this point (mx,my) are full-pell and the relative displacement */
1114 ppix = c->ref[0][0] + (my * s->linesize) + mx;
1116 vard = s->dsp.sse[0](NULL, pix, ppix, s->linesize, 16);
1118 pic->mc_mb_var[s->mb_stride * mb_y + mb_x] = (vard+128)>>8;
1119 // pic->mb_cmp_score[s->mb_stride * mb_y + mb_x] = dmin;
1120 c->mc_mb_var_sum_temp += (vard+128)>>8;
1123 printf("varc=%4d avg_var=%4d (sum=%4d) vard=%4d mx=%2d my=%2d\n",
1124 varc, s->avg_mb_var, sum, vard, mx - xx, my - yy);
1127 int p_score= FFMIN(vard, varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*100);
1128 int i_score= varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*20;
1129 c->scene_change_score+= ff_sqrt(p_score) - ff_sqrt(i_score);
1131 if(mb_type == CANDIDATE_MB_TYPE_INTER){
1132 c->sub_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
1133 set_p_mv_tables(s, mx, my, 1);
1138 if(mb_type == CANDIDATE_MB_TYPE_INTER4V){
1139 h263_mv4_search(s, mx, my, shift);
1141 set_p_mv_tables(s, mx, my, 0);
1143 if(mb_type == CANDIDATE_MB_TYPE_INTER_I){
1144 interlaced_search(s, 0, s->p_field_mv_table, s->p_field_select_table, mx, my, 1);
1146 }else if(c->avctx->mb_decision > FF_MB_DECISION_SIMPLE){
1147 int p_score= FFMIN(vard, varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*100);
1148 int i_score= varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*20;
1149 c->scene_change_score+= ff_sqrt(p_score) - ff_sqrt(i_score);
1151 if (vard*2 + 200*256 > varc)
1152 mb_type|= CANDIDATE_MB_TYPE_INTRA;
1153 if (varc*2 + 200*256 > vard || s->qscale > 24){
1154 // if (varc*2 + 200*256 + 50*(s->lambda2>>FF_LAMBDA_SHIFT) > vard){
1155 mb_type|= CANDIDATE_MB_TYPE_INTER;
1156 c->sub_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
1157 if(s->flags&CODEC_FLAG_MV0)
1159 mb_type |= CANDIDATE_MB_TYPE_SKIPPED; //FIXME check difference
1164 if((s->flags&CODEC_FLAG_4MV)
1165 && !c->skip && varc>50<<8 && vard>10<<8){
1166 if(h263_mv4_search(s, mx, my, shift) < INT_MAX)
1167 mb_type|=CANDIDATE_MB_TYPE_INTER4V;
1169 set_p_mv_tables(s, mx, my, 0);
1171 set_p_mv_tables(s, mx, my, 1);
1172 if((s->flags&CODEC_FLAG_INTERLACED_ME)
1173 && !c->skip){ //FIXME varc/d checks
1174 if(interlaced_search(s, 0, s->p_field_mv_table, s->p_field_select_table, mx, my, 0) < INT_MAX)
1175 mb_type |= CANDIDATE_MB_TYPE_INTER_I;
1179 mb_type= CANDIDATE_MB_TYPE_INTER;
1181 dmin= c->sub_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
1182 if(c->avctx->me_sub_cmp != c->avctx->mb_cmp && !c->skip)
1183 dmin= ff_get_mb_score(s, mx, my, 0, 0, 0, 16, 1);
1185 if((s->flags&CODEC_FLAG_4MV)
1186 && !c->skip && varc>50<<8 && vard>10<<8){
1187 int dmin4= h263_mv4_search(s, mx, my, shift);
1189 mb_type= CANDIDATE_MB_TYPE_INTER4V;
1193 if((s->flags&CODEC_FLAG_INTERLACED_ME)
1194 && !c->skip){ //FIXME varc/d checks
1195 int dmin_i= interlaced_search(s, 0, s->p_field_mv_table, s->p_field_select_table, mx, my, 0);
1197 mb_type = CANDIDATE_MB_TYPE_INTER_I;
1202 // pic->mb_cmp_score[s->mb_stride * mb_y + mb_x] = dmin;
1203 set_p_mv_tables(s, mx, my, mb_type!=CANDIDATE_MB_TYPE_INTER4V);
1205 /* get intra luma score */
1206 if((c->avctx->mb_cmp&0xFF)==FF_CMP_SSE){
1207 intra_score= varc - 500;
1209 int mean= (sum+128)>>8;
1212 for(i=0; i<16; i++){
1213 *(uint32_t*)(&c->scratchpad[i*s->linesize+ 0]) = mean;
1214 *(uint32_t*)(&c->scratchpad[i*s->linesize+ 4]) = mean;
1215 *(uint32_t*)(&c->scratchpad[i*s->linesize+ 8]) = mean;
1216 *(uint32_t*)(&c->scratchpad[i*s->linesize+12]) = mean;
1219 intra_score= s->dsp.mb_cmp[0](s, c->scratchpad, pix, s->linesize, 16);
1222 /* get chroma score */
1223 if(c->avctx->mb_cmp&FF_CMP_CHROMA){
1228 if(s->out_format == FMT_H263){
1229 mean= (s->dc_val[i][mb_x + mb_y*s->b8_stride] + 4)>>3; //FIXME not exact but simple ;)
1231 mean= (s->last_dc[i] + 4)>>3;
1233 dest_c = s->new_picture.data[i] + (mb_y * 8 * (s->uvlinesize)) + mb_x * 8;
1237 *(uint32_t*)(&c->scratchpad[i*s->uvlinesize+ 0]) = mean;
1238 *(uint32_t*)(&c->scratchpad[i*s->uvlinesize+ 4]) = mean;
1241 intra_score+= s->dsp.mb_cmp[1](s, c->scratchpad, dest_c, s->uvlinesize);
1245 intra_score += c->mb_penalty_factor*16;
1247 if(intra_score < dmin){
1248 mb_type= CANDIDATE_MB_TYPE_INTRA;
1249 s->current_picture.mb_type[mb_y*s->mb_stride + mb_x]= CANDIDATE_MB_TYPE_INTRA; //FIXME cleanup
1251 s->current_picture.mb_type[mb_y*s->mb_stride + mb_x]= 0;
1254 int p_score= FFMIN(vard, varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*100);
1255 int i_score= varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*20;
1256 c->scene_change_score+= ff_sqrt(p_score) - ff_sqrt(i_score);
1260 s->mb_type[mb_y*s->mb_stride + mb_x]= mb_type;
1263 int ff_pre_estimate_p_frame_motion(MpegEncContext * s,
1266 MotionEstContext * const c= &s->me;
1269 const int shift= 1+s->quarter_sample;
1270 const int xy= mb_x + mb_y*s->mb_stride;
1271 init_ref(c, s->new_picture.data, s->last_picture.data, NULL, 16*mb_x, 16*mb_y, 0);
1273 assert(s->quarter_sample==0 || s->quarter_sample==1);
1275 c->pre_penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_pre_cmp);
1276 c->current_mv_penalty= c->mv_penalty[s->f_code] + MAX_MV;
1278 get_limits(s, 16*mb_x, 16*mb_y);
1281 P_LEFT[0] = s->p_mv_table[xy + 1][0];
1282 P_LEFT[1] = s->p_mv_table[xy + 1][1];
1284 if(P_LEFT[0] < (c->xmin<<shift)) P_LEFT[0] = (c->xmin<<shift);
1286 /* special case for first line */
1287 if (s->first_slice_line) {
1288 c->pred_x= P_LEFT[0];
1289 c->pred_y= P_LEFT[1];
1290 P_TOP[0]= P_TOPRIGHT[0]= P_MEDIAN[0]=
1291 P_TOP[1]= P_TOPRIGHT[1]= P_MEDIAN[1]= 0; //FIXME
1293 P_TOP[0] = s->p_mv_table[xy + s->mb_stride ][0];
1294 P_TOP[1] = s->p_mv_table[xy + s->mb_stride ][1];
1295 P_TOPRIGHT[0] = s->p_mv_table[xy + s->mb_stride - 1][0];
1296 P_TOPRIGHT[1] = s->p_mv_table[xy + s->mb_stride - 1][1];
1297 if(P_TOP[1] < (c->ymin<<shift)) P_TOP[1] = (c->ymin<<shift);
1298 if(P_TOPRIGHT[0] > (c->xmax<<shift)) P_TOPRIGHT[0]= (c->xmax<<shift);
1299 if(P_TOPRIGHT[1] < (c->ymin<<shift)) P_TOPRIGHT[1]= (c->ymin<<shift);
1301 P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
1302 P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
1304 c->pred_x = P_MEDIAN[0];
1305 c->pred_y = P_MEDIAN[1];
1308 dmin = ff_epzs_motion_search(s, &mx, &my, P, 0, 0, s->p_mv_table, (1<<16)>>shift, 0, 16);
1310 s->p_mv_table[xy][0] = mx<<shift;
1311 s->p_mv_table[xy][1] = my<<shift;
1316 static int ff_estimate_motion_b(MpegEncContext * s,
1317 int mb_x, int mb_y, int16_t (*mv_table)[2], int ref_index, int f_code)
1319 MotionEstContext * const c= &s->me;
1322 const int shift= 1+s->quarter_sample;
1323 const int mot_stride = s->mb_stride;
1324 const int mot_xy = mb_y*mot_stride + mb_x;
1325 uint8_t * const mv_penalty= c->mv_penalty[f_code] + MAX_MV;
1328 c->penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_cmp);
1329 c->sub_penalty_factor= get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_sub_cmp);
1330 c->mb_penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->mb_cmp);
1331 c->current_mv_penalty= mv_penalty;
1333 get_limits(s, 16*mb_x, 16*mb_y);
1335 switch(s->me_method) {
1338 no_motion_search(s, &mx, &my);
1346 P_LEFT[0] = mv_table[mot_xy - 1][0];
1347 P_LEFT[1] = mv_table[mot_xy - 1][1];
1349 if(P_LEFT[0] > (c->xmax<<shift)) P_LEFT[0] = (c->xmax<<shift);
1351 /* special case for first line */
1352 if (!s->first_slice_line) {
1353 P_TOP[0] = mv_table[mot_xy - mot_stride ][0];
1354 P_TOP[1] = mv_table[mot_xy - mot_stride ][1];
1355 P_TOPRIGHT[0] = mv_table[mot_xy - mot_stride + 1 ][0];
1356 P_TOPRIGHT[1] = mv_table[mot_xy - mot_stride + 1 ][1];
1357 if(P_TOP[1] > (c->ymax<<shift)) P_TOP[1]= (c->ymax<<shift);
1358 if(P_TOPRIGHT[0] < (c->xmin<<shift)) P_TOPRIGHT[0]= (c->xmin<<shift);
1359 if(P_TOPRIGHT[1] > (c->ymax<<shift)) P_TOPRIGHT[1]= (c->ymax<<shift);
1361 P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
1362 P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
1364 c->pred_x= P_LEFT[0];
1365 c->pred_y= P_LEFT[1];
1368 if(mv_table == s->b_forw_mv_table){
1369 mv_scale= (s->pb_time<<16) / (s->pp_time<<shift);
1371 mv_scale= ((s->pb_time - s->pp_time)<<16) / (s->pp_time<<shift);
1374 dmin = ff_epzs_motion_search(s, &mx, &my, P, 0, ref_index, s->p_mv_table, mv_scale, 0, 16);
1379 dmin= c->sub_motion_search(s, &mx, &my, dmin, 0, ref_index, 0, 16);
1381 if(c->avctx->me_sub_cmp != c->avctx->mb_cmp && !c->skip)
1382 dmin= ff_get_mb_score(s, mx, my, 0, ref_index, 0, 16, 1);
1384 //printf("%d %d %d %d//", s->mb_x, s->mb_y, mx, my);
1385 // s->mb_type[mb_y*s->mb_width + mb_x]= mb_type;
1386 mv_table[mot_xy][0]= mx;
1387 mv_table[mot_xy][1]= my;
1392 static inline int check_bidir_mv(MpegEncContext * s,
1393 int motion_fx, int motion_fy,
1394 int motion_bx, int motion_by,
1395 int pred_fx, int pred_fy,
1396 int pred_bx, int pred_by,
1400 //FIXME better f_code prediction (max mv & distance)
1402 MotionEstContext * const c= &s->me;
1403 uint8_t * const mv_penalty_f= c->mv_penalty[s->f_code] + MAX_MV; // f_code of the prev frame
1404 uint8_t * const mv_penalty_b= c->mv_penalty[s->b_code] + MAX_MV; // f_code of the prev frame
1405 int stride= c->stride;
1406 uint8_t *dest_y = c->scratchpad;
1411 uint8_t **src_data= c->src[0];
1412 uint8_t **ref_data= c->ref[0];
1413 uint8_t **ref2_data= c->ref[2];
1415 if(s->quarter_sample){
1416 dxy = ((motion_fy & 3) << 2) | (motion_fx & 3);
1417 src_x = motion_fx >> 2;
1418 src_y = motion_fy >> 2;
1420 ptr = ref_data[0] + (src_y * stride) + src_x;
1421 s->dsp.put_qpel_pixels_tab[0][dxy](dest_y , ptr , stride);
1423 dxy = ((motion_by & 3) << 2) | (motion_bx & 3);
1424 src_x = motion_bx >> 2;
1425 src_y = motion_by >> 2;
1427 ptr = ref2_data[0] + (src_y * stride) + src_x;
1428 s->dsp.avg_qpel_pixels_tab[size][dxy](dest_y , ptr , stride);
1430 dxy = ((motion_fy & 1) << 1) | (motion_fx & 1);
1431 src_x = motion_fx >> 1;
1432 src_y = motion_fy >> 1;
1434 ptr = ref_data[0] + (src_y * stride) + src_x;
1435 s->dsp.put_pixels_tab[size][dxy](dest_y , ptr , stride, h);
1437 dxy = ((motion_by & 1) << 1) | (motion_bx & 1);
1438 src_x = motion_bx >> 1;
1439 src_y = motion_by >> 1;
1441 ptr = ref2_data[0] + (src_y * stride) + src_x;
1442 s->dsp.avg_pixels_tab[size][dxy](dest_y , ptr , stride, h);
1445 fbmin = (mv_penalty_f[motion_fx-pred_fx] + mv_penalty_f[motion_fy-pred_fy])*c->mb_penalty_factor
1446 +(mv_penalty_b[motion_bx-pred_bx] + mv_penalty_b[motion_by-pred_by])*c->mb_penalty_factor
1447 + s->dsp.mb_cmp[size](s, src_data[0], dest_y, stride, h); //FIXME new_pic
1449 if(c->avctx->mb_cmp&FF_CMP_CHROMA){
1456 /* refine the bidir vectors in hq mode and return the score in both lq & hq mode*/
1457 static inline int bidir_refine(MpegEncContext * s, int mb_x, int mb_y)
1459 MotionEstContext * const c= &s->me;
1460 const int mot_stride = s->mb_stride;
1461 const int xy = mb_y *mot_stride + mb_x;
1463 int pred_fx= s->b_bidir_forw_mv_table[xy-1][0];
1464 int pred_fy= s->b_bidir_forw_mv_table[xy-1][1];
1465 int pred_bx= s->b_bidir_back_mv_table[xy-1][0];
1466 int pred_by= s->b_bidir_back_mv_table[xy-1][1];
1467 int motion_fx= s->b_bidir_forw_mv_table[xy][0]= s->b_forw_mv_table[xy][0];
1468 int motion_fy= s->b_bidir_forw_mv_table[xy][1]= s->b_forw_mv_table[xy][1];
1469 int motion_bx= s->b_bidir_back_mv_table[xy][0]= s->b_back_mv_table[xy][0];
1470 int motion_by= s->b_bidir_back_mv_table[xy][1]= s->b_back_mv_table[xy][1];
1471 const int flags= c->sub_flags;
1472 const int qpel= flags&FLAG_QPEL;
1473 const int shift= 1+qpel;
1474 const int xmin= c->xmin<<shift;
1475 const int ymin= c->ymin<<shift;
1476 const int xmax= c->xmax<<shift;
1477 const int ymax= c->ymax<<shift;
1478 #define HASH(fx,fy,bx,by) ((fx)+17*(fy)+63*(bx)+117*(by))
1479 int hashidx= HASH(motion_fx,motion_fy, motion_bx, motion_by);
1482 memset(map,0,sizeof(map));
1483 map[hashidx&255] = 1;
1485 fbmin= check_bidir_mv(s, motion_fx, motion_fy,
1486 motion_bx, motion_by,
1491 if(s->avctx->bidir_refine){
1493 static const uint8_t limittab[5]={0,8,32,64,80};
1494 const int limit= limittab[s->avctx->bidir_refine];
1495 static const int8_t vect[][4]={
1496 { 0, 0, 0, 1}, { 0, 0, 0,-1}, { 0, 0, 1, 0}, { 0, 0,-1, 0}, { 0, 1, 0, 0}, { 0,-1, 0, 0}, { 1, 0, 0, 0}, {-1, 0, 0, 0},
1498 { 0, 0, 1, 1}, { 0, 0,-1,-1}, { 0, 1, 1, 0}, { 0,-1,-1, 0}, { 1, 1, 0, 0}, {-1,-1, 0, 0}, { 1, 0, 0, 1}, {-1, 0, 0,-1},
1499 { 0, 1, 0, 1}, { 0,-1, 0,-1}, { 1, 0, 1, 0}, {-1, 0,-1, 0},
1500 { 0, 0,-1, 1}, { 0, 0, 1,-1}, { 0,-1, 1, 0}, { 0, 1,-1, 0}, {-1, 1, 0, 0}, { 1,-1, 0, 0}, { 1, 0, 0,-1}, {-1, 0, 0, 1},
1501 { 0,-1, 0, 1}, { 0, 1, 0,-1}, {-1, 0, 1, 0}, { 1, 0,-1, 0},
1503 { 0, 1, 1, 1}, { 0,-1,-1,-1}, { 1, 1, 1, 0}, {-1,-1,-1, 0}, { 1, 1, 0, 1}, {-1,-1, 0,-1}, { 1, 0, 1, 1}, {-1, 0,-1,-1},
1504 { 0,-1, 1, 1}, { 0, 1,-1,-1}, {-1, 1, 1, 0}, { 1,-1,-1, 0}, { 1, 1, 0,-1}, {-1,-1, 0, 1}, { 1, 0,-1, 1}, {-1, 0, 1,-1},
1505 { 0, 1,-1, 1}, { 0,-1, 1,-1}, { 1,-1, 1, 0}, {-1, 1,-1, 0}, {-1, 1, 0, 1}, { 1,-1, 0,-1}, { 1, 0, 1,-1}, {-1, 0,-1, 1},
1506 { 0, 1, 1,-1}, { 0,-1,-1, 1}, { 1, 1,-1, 0}, {-1,-1, 1, 0}, { 1,-1, 0, 1}, {-1, 1, 0,-1}, {-1, 0, 1, 1}, { 1, 0,-1,-1},
1508 { 1, 1, 1, 1}, {-1,-1,-1,-1},
1509 { 1, 1, 1,-1}, {-1,-1,-1, 1}, { 1, 1,-1, 1}, {-1,-1, 1,-1}, { 1,-1, 1, 1}, {-1, 1,-1,-1}, {-1, 1, 1, 1}, { 1,-1,-1,-1},
1510 { 1, 1,-1,-1}, {-1,-1, 1, 1}, { 1,-1,-1, 1}, {-1, 1, 1,-1}, { 1,-1, 1,-1}, {-1, 1,-1, 1},
1512 static const uint8_t hash[]={
1513 HASH( 0, 0, 0, 1), HASH( 0, 0, 0,-1), HASH( 0, 0, 1, 0), HASH( 0, 0,-1, 0), HASH( 0, 1, 0, 0), HASH( 0,-1, 0, 0), HASH( 1, 0, 0, 0), HASH(-1, 0, 0, 0),
1515 HASH( 0, 0, 1, 1), HASH( 0, 0,-1,-1), HASH( 0, 1, 1, 0), HASH( 0,-1,-1, 0), HASH( 1, 1, 0, 0), HASH(-1,-1, 0, 0), HASH( 1, 0, 0, 1), HASH(-1, 0, 0,-1),
1516 HASH( 0, 1, 0, 1), HASH( 0,-1, 0,-1), HASH( 1, 0, 1, 0), HASH(-1, 0,-1, 0),
1517 HASH( 0, 0,-1, 1), HASH( 0, 0, 1,-1), HASH( 0,-1, 1, 0), HASH( 0, 1,-1, 0), HASH(-1, 1, 0, 0), HASH( 1,-1, 0, 0), HASH( 1, 0, 0,-1), HASH(-1, 0, 0, 1),
1518 HASH( 0,-1, 0, 1), HASH( 0, 1, 0,-1), HASH(-1, 0, 1, 0), HASH( 1, 0,-1, 0),
1520 HASH( 0, 1, 1, 1), HASH( 0,-1,-1,-1), HASH( 1, 1, 1, 0), HASH(-1,-1,-1, 0), HASH( 1, 1, 0, 1), HASH(-1,-1, 0,-1), HASH( 1, 0, 1, 1), HASH(-1, 0,-1,-1),
1521 HASH( 0,-1, 1, 1), HASH( 0, 1,-1,-1), HASH(-1, 1, 1, 0), HASH( 1,-1,-1, 0), HASH( 1, 1, 0,-1), HASH(-1,-1, 0, 1), HASH( 1, 0,-1, 1), HASH(-1, 0, 1,-1),
1522 HASH( 0, 1,-1, 1), HASH( 0,-1, 1,-1), HASH( 1,-1, 1, 0), HASH(-1, 1,-1, 0), HASH(-1, 1, 0, 1), HASH( 1,-1, 0,-1), HASH( 1, 0, 1,-1), HASH(-1, 0,-1, 1),
1523 HASH( 0, 1, 1,-1), HASH( 0,-1,-1, 1), HASH( 1, 1,-1, 0), HASH(-1,-1, 1, 0), HASH( 1,-1, 0, 1), HASH(-1, 1, 0,-1), HASH(-1, 0, 1, 1), HASH( 1, 0,-1,-1),
1525 HASH( 1, 1, 1, 1), HASH(-1,-1,-1,-1),
1526 HASH( 1, 1, 1,-1), HASH(-1,-1,-1, 1), HASH( 1, 1,-1, 1), HASH(-1,-1, 1,-1), HASH( 1,-1, 1, 1), HASH(-1, 1,-1,-1), HASH(-1, 1, 1, 1), HASH( 1,-1,-1,-1),
1527 HASH( 1, 1,-1,-1), HASH(-1,-1, 1, 1), HASH( 1,-1,-1, 1), HASH(-1, 1, 1,-1), HASH( 1,-1, 1,-1), HASH(-1, 1,-1, 1),
1530 #define CHECK_BIDIR(fx,fy,bx,by)\
1531 if( !map[(hashidx+HASH(fx,fy,bx,by))&255]\
1532 &&(fx<=0 || motion_fx+fx<=xmax) && (fy<=0 || motion_fy+fy<=ymax) && (bx<=0 || motion_bx+bx<=xmax) && (by<=0 || motion_by+by<=ymax)\
1533 &&(fx>=0 || motion_fx+fx>=xmin) && (fy>=0 || motion_fy+fy>=ymin) && (bx>=0 || motion_bx+bx>=xmin) && (by>=0 || motion_by+by>=ymin)){\
1535 map[(hashidx+HASH(fx,fy,bx,by))&255] = 1;\
1536 score= check_bidir_mv(s, motion_fx+fx, motion_fy+fy, motion_bx+bx, motion_by+by, pred_fx, pred_fy, pred_bx, pred_by, 0, 16);\
1538 hashidx += HASH(fx,fy,bx,by);\
1547 #define CHECK_BIDIR2(a,b,c,d)\
1548 CHECK_BIDIR(a,b,c,d)\
1549 CHECK_BIDIR(-(a),-(b),-(c),-(d))
1556 CHECK_BIDIR2(0,0,0,1)
1557 CHECK_BIDIR2(0,0,1,0)
1558 CHECK_BIDIR2(0,1,0,0)
1559 CHECK_BIDIR2(1,0,0,0)
1561 for(i=8; i<limit; i++){
1562 int fx= motion_fx+vect[i][0];
1563 int fy= motion_fy+vect[i][1];
1564 int bx= motion_bx+vect[i][2];
1565 int by= motion_by+vect[i][3];
1567 int a= (xmax - FFMAX(fx,bx))|(FFMIN(fx,bx) - xmin);
1568 int b= (ymax - FFMAX(fy,by))|(FFMIN(fy,by) - ymin);
1570 map[(hashidx+hash[i])&255] = 1;
1572 if(!map[(hashidx+hash[i])&255]){
1574 map[(hashidx+hash[i])&255] = 1;
1575 score= check_bidir_mv(s, fx, fy, bx, by, pred_fx, pred_fy, pred_bx, pred_by, 0, 16);
1586 int a= FFMIN(xmax - FFMAX(fx,bx), FFMIN(fx,bx) - xmin);
1587 int b= FFMIN(ymax - FFMAX(fy,by), FFMIN(fy,by) - ymin);
1588 borderdist= FFMIN(a,b);
1596 s->b_bidir_forw_mv_table[xy][0]= motion_fx;
1597 s->b_bidir_forw_mv_table[xy][1]= motion_fy;
1598 s->b_bidir_back_mv_table[xy][0]= motion_bx;
1599 s->b_bidir_back_mv_table[xy][1]= motion_by;
1604 static inline int direct_search(MpegEncContext * s, int mb_x, int mb_y)
1606 MotionEstContext * const c= &s->me;
1608 const int mot_stride = s->mb_stride;
1609 const int mot_xy = mb_y*mot_stride + mb_x;
1610 const int shift= 1+s->quarter_sample;
1612 const int time_pp= s->pp_time;
1613 const int time_pb= s->pb_time;
1614 int mx, my, xmin, xmax, ymin, ymax;
1615 int16_t (*mv_table)[2]= s->b_direct_mv_table;
1617 c->current_mv_penalty= c->mv_penalty[1] + MAX_MV;
1618 ymin= xmin=(-32)>>shift;
1619 ymax= xmax= 31>>shift;
1621 if(IS_8X8(s->next_picture.mb_type[mot_xy])){
1622 s->mv_type= MV_TYPE_8X8;
1624 s->mv_type= MV_TYPE_16X16;
1628 int index= s->block_index[i];
1631 c->co_located_mv[i][0]= s->next_picture.motion_val[0][index][0];
1632 c->co_located_mv[i][1]= s->next_picture.motion_val[0][index][1];
1633 c->direct_basis_mv[i][0]= c->co_located_mv[i][0]*time_pb/time_pp + ((i& 1)<<(shift+3));
1634 c->direct_basis_mv[i][1]= c->co_located_mv[i][1]*time_pb/time_pp + ((i>>1)<<(shift+3));
1635 // c->direct_basis_mv[1][i][0]= c->co_located_mv[i][0]*(time_pb - time_pp)/time_pp + ((i &1)<<(shift+3);
1636 // c->direct_basis_mv[1][i][1]= c->co_located_mv[i][1]*(time_pb - time_pp)/time_pp + ((i>>1)<<(shift+3);
1638 max= FFMAX(c->direct_basis_mv[i][0], c->direct_basis_mv[i][0] - c->co_located_mv[i][0])>>shift;
1639 min= FFMIN(c->direct_basis_mv[i][0], c->direct_basis_mv[i][0] - c->co_located_mv[i][0])>>shift;
1640 max+= 16*mb_x + 1; // +-1 is for the simpler rounding
1642 xmax= FFMIN(xmax, s->width - max);
1643 xmin= FFMAX(xmin, - 16 - min);
1645 max= FFMAX(c->direct_basis_mv[i][1], c->direct_basis_mv[i][1] - c->co_located_mv[i][1])>>shift;
1646 min= FFMIN(c->direct_basis_mv[i][1], c->direct_basis_mv[i][1] - c->co_located_mv[i][1])>>shift;
1647 max+= 16*mb_y + 1; // +-1 is for the simpler rounding
1649 ymax= FFMIN(ymax, s->height - max);
1650 ymin= FFMAX(ymin, - 16 - min);
1652 if(s->mv_type == MV_TYPE_16X16) break;
1655 assert(xmax <= 15 && ymax <= 15 && xmin >= -16 && ymin >= -16);
1657 if(xmax < 0 || xmin >0 || ymax < 0 || ymin > 0){
1658 s->b_direct_mv_table[mot_xy][0]= 0;
1659 s->b_direct_mv_table[mot_xy][1]= 0;
1661 return 256*256*256*64;
1668 c->flags |= FLAG_DIRECT;
1669 c->sub_flags |= FLAG_DIRECT;
1673 P_LEFT[0] = av_clip(mv_table[mot_xy - 1][0], xmin<<shift, xmax<<shift);
1674 P_LEFT[1] = av_clip(mv_table[mot_xy - 1][1], ymin<<shift, ymax<<shift);
1676 /* special case for first line */
1677 if (!s->first_slice_line) { //FIXME maybe allow this over thread boundary as it is clipped
1678 P_TOP[0] = av_clip(mv_table[mot_xy - mot_stride ][0], xmin<<shift, xmax<<shift);
1679 P_TOP[1] = av_clip(mv_table[mot_xy - mot_stride ][1], ymin<<shift, ymax<<shift);
1680 P_TOPRIGHT[0] = av_clip(mv_table[mot_xy - mot_stride + 1 ][0], xmin<<shift, xmax<<shift);
1681 P_TOPRIGHT[1] = av_clip(mv_table[mot_xy - mot_stride + 1 ][1], ymin<<shift, ymax<<shift);
1683 P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
1684 P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
1687 dmin = ff_epzs_motion_search(s, &mx, &my, P, 0, 0, mv_table, 1<<(16-shift), 0, 16);
1688 if(c->sub_flags&FLAG_QPEL)
1689 dmin = qpel_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
1691 dmin = hpel_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
1693 if(c->avctx->me_sub_cmp != c->avctx->mb_cmp && !c->skip)
1694 dmin= ff_get_mb_score(s, mx, my, 0, 0, 0, 16, 1);
1696 get_limits(s, 16*mb_x, 16*mb_y); //restore c->?min/max, maybe not needed
1698 mv_table[mot_xy][0]= mx;
1699 mv_table[mot_xy][1]= my;
1700 c->flags &= ~FLAG_DIRECT;
1701 c->sub_flags &= ~FLAG_DIRECT;
1706 void ff_estimate_b_frame_motion(MpegEncContext * s,
1709 MotionEstContext * const c= &s->me;
1710 const int penalty_factor= c->mb_penalty_factor;
1711 int fmin, bmin, dmin, fbmin, bimin, fimin;
1713 const int xy = mb_y*s->mb_stride + mb_x;
1714 init_ref(c, s->new_picture.data, s->last_picture.data, s->next_picture.data, 16*mb_x, 16*mb_y, 2);
1716 get_limits(s, 16*mb_x, 16*mb_y);
1720 if(s->codec_id == CODEC_ID_MPEG4 && s->next_picture.mbskip_table[xy]){
1721 int score= direct_search(s, mb_x, mb_y); //FIXME just check 0,0
1723 score= ((unsigned)(score*score + 128*256))>>16;
1724 c->mc_mb_var_sum_temp += score;
1725 s->current_picture.mc_mb_var[mb_y*s->mb_stride + mb_x] = score; //FIXME use SSE
1726 s->mb_type[mb_y*s->mb_stride + mb_x]= CANDIDATE_MB_TYPE_DIRECT0;
1731 if(c->avctx->me_threshold){
1732 int vard= check_input_motion(s, mb_x, mb_y, 0);
1734 if((vard+128)>>8 < c->avctx->me_threshold){
1735 // pix = c->src[0][0];
1736 // sum = s->dsp.pix_sum(pix, s->linesize);
1737 // varc = s->dsp.pix_norm1(pix, s->linesize) - (((unsigned)(sum*sum))>>8) + 500;
1739 // pic->mb_var [s->mb_stride * mb_y + mb_x] = (varc+128)>>8;
1740 s->current_picture.mc_mb_var[s->mb_stride * mb_y + mb_x] = (vard+128)>>8;
1741 /* pic->mb_mean [s->mb_stride * mb_y + mb_x] = (sum+128)>>8;
1742 c->mb_var_sum_temp += (varc+128)>>8;*/
1743 c->mc_mb_var_sum_temp += (vard+128)>>8;
1744 /* if (vard <= 64<<8 || vard < varc) {
1745 c->scene_change_score+= ff_sqrt(vard) - ff_sqrt(varc);
1747 c->scene_change_score+= s->qscale * s->avctx->scenechange_factor;
1751 if((vard+128)>>8 < c->avctx->mb_threshold){
1752 type= s->mb_type[mb_y*s->mb_stride + mb_x];
1753 if(type == CANDIDATE_MB_TYPE_DIRECT){
1754 direct_search(s, mb_x, mb_y);
1756 if(type == CANDIDATE_MB_TYPE_FORWARD || type == CANDIDATE_MB_TYPE_BIDIR){
1758 ff_estimate_motion_b(s, mb_x, mb_y, s->b_forw_mv_table, 0, s->f_code);
1760 if(type == CANDIDATE_MB_TYPE_BACKWARD || type == CANDIDATE_MB_TYPE_BIDIR){
1762 ff_estimate_motion_b(s, mb_x, mb_y, s->b_back_mv_table, 2, s->b_code);
1764 if(type == CANDIDATE_MB_TYPE_FORWARD_I || type == CANDIDATE_MB_TYPE_BIDIR_I){
1766 c->current_mv_penalty= c->mv_penalty[s->f_code] + MAX_MV;
1767 interlaced_search(s, 0,
1768 s->b_field_mv_table[0], s->b_field_select_table[0],
1769 s->b_forw_mv_table[xy][0], s->b_forw_mv_table[xy][1], 1);
1771 if(type == CANDIDATE_MB_TYPE_BACKWARD_I || type == CANDIDATE_MB_TYPE_BIDIR_I){
1773 c->current_mv_penalty= c->mv_penalty[s->b_code] + MAX_MV;
1774 interlaced_search(s, 2,
1775 s->b_field_mv_table[1], s->b_field_select_table[1],
1776 s->b_back_mv_table[xy][0], s->b_back_mv_table[xy][1], 1);
1782 if (s->codec_id == CODEC_ID_MPEG4)
1783 dmin= direct_search(s, mb_x, mb_y);
1786 //FIXME penalty stuff for non mpeg4
1788 fmin= ff_estimate_motion_b(s, mb_x, mb_y, s->b_forw_mv_table, 0, s->f_code) + 3*penalty_factor;
1791 bmin= ff_estimate_motion_b(s, mb_x, mb_y, s->b_back_mv_table, 2, s->b_code) + 2*penalty_factor;
1792 //printf(" %d %d ", s->b_forw_mv_table[xy][0], s->b_forw_mv_table[xy][1]);
1795 fbmin= bidir_refine(s, mb_x, mb_y) + penalty_factor;
1796 //printf("%d %d %d %d\n", dmin, fmin, bmin, fbmin);
1798 if(s->flags & CODEC_FLAG_INTERLACED_ME){
1799 //FIXME mb type penalty
1801 c->current_mv_penalty= c->mv_penalty[s->f_code] + MAX_MV;
1802 fimin= interlaced_search(s, 0,
1803 s->b_field_mv_table[0], s->b_field_select_table[0],
1804 s->b_forw_mv_table[xy][0], s->b_forw_mv_table[xy][1], 0);
1805 c->current_mv_penalty= c->mv_penalty[s->b_code] + MAX_MV;
1806 bimin= interlaced_search(s, 2,
1807 s->b_field_mv_table[1], s->b_field_select_table[1],
1808 s->b_back_mv_table[xy][0], s->b_back_mv_table[xy][1], 0);
1810 fimin= bimin= INT_MAX;
1814 type = CANDIDATE_MB_TYPE_FORWARD;
1818 type = CANDIDATE_MB_TYPE_DIRECT;
1822 type= CANDIDATE_MB_TYPE_BACKWARD;
1826 type= CANDIDATE_MB_TYPE_BIDIR;
1830 type= CANDIDATE_MB_TYPE_FORWARD_I;
1834 type= CANDIDATE_MB_TYPE_BACKWARD_I;
1837 score= ((unsigned)(score*score + 128*256))>>16;
1838 c->mc_mb_var_sum_temp += score;
1839 s->current_picture.mc_mb_var[mb_y*s->mb_stride + mb_x] = score; //FIXME use SSE
1842 if(c->avctx->mb_decision > FF_MB_DECISION_SIMPLE){
1843 type= CANDIDATE_MB_TYPE_FORWARD | CANDIDATE_MB_TYPE_BACKWARD | CANDIDATE_MB_TYPE_BIDIR | CANDIDATE_MB_TYPE_DIRECT;
1845 type |= CANDIDATE_MB_TYPE_FORWARD_I;
1847 type |= CANDIDATE_MB_TYPE_BACKWARD_I;
1848 if(fimin < INT_MAX && bimin < INT_MAX){
1849 type |= CANDIDATE_MB_TYPE_BIDIR_I;
1851 //FIXME something smarter
1852 if(dmin>256*256*16) type&= ~CANDIDATE_MB_TYPE_DIRECT; //do not try direct mode if it is invalid for this MB
1853 if(s->codec_id == CODEC_ID_MPEG4 && type&CANDIDATE_MB_TYPE_DIRECT && s->flags&CODEC_FLAG_MV0 && *(uint32_t*)s->b_direct_mv_table[xy])
1854 type |= CANDIDATE_MB_TYPE_DIRECT0;
1856 if(s->out_format == FMT_MPEG1)
1857 type |= CANDIDATE_MB_TYPE_INTRA;
1861 s->mb_type[mb_y*s->mb_stride + mb_x]= type;
1864 /* find best f_code for ME which do unlimited searches */
1865 int ff_get_best_fcode(MpegEncContext * s, int16_t (*mv_table)[2], int type)
1867 if(s->me_method>=ME_EPZS){
1869 int i, y, range= s->avctx->me_range ? s->avctx->me_range : (INT_MAX/2);
1870 uint8_t * fcode_tab= s->fcode_tab;
1872 int best_score=-10000000;
1874 if(s->msmpeg4_version)
1875 range= FFMIN(range, 16);
1876 else if(s->codec_id == CODEC_ID_MPEG2VIDEO && s->avctx->strict_std_compliance >= FF_COMPLIANCE_NORMAL)
1877 range= FFMIN(range, 256);
1879 for(i=0; i<8; i++) score[i]= s->mb_num*(8-i);
1881 for(y=0; y<s->mb_height; y++){
1883 int xy= y*s->mb_stride;
1884 for(x=0; x<s->mb_width; x++){
1885 if(s->mb_type[xy] & type){
1886 int mx= mv_table[xy][0];
1887 int my= mv_table[xy][1];
1888 int fcode= FFMAX(fcode_tab[mx + MAX_MV],
1889 fcode_tab[my + MAX_MV]);
1892 if(mx >= range || mx < -range ||
1893 my >= range || my < -range)
1896 for(j=0; j<fcode && j<8; j++){
1897 if(s->pict_type==FF_B_TYPE || s->current_picture.mc_mb_var[xy] < s->current_picture.mb_var[xy])
1906 if(score[i] > best_score){
1907 best_score= score[i];
1910 // printf("%d %d\n", i, score[i]);
1913 // printf("fcode: %d type: %d\n", i, s->pict_type);
1915 /* for(i=0; i<=MAX_FCODE; i++){
1916 printf("%d ", mv_num[i]);
1924 void ff_fix_long_p_mvs(MpegEncContext * s)
1926 MotionEstContext * const c= &s->me;
1927 const int f_code= s->f_code;
1929 assert(s->pict_type==FF_P_TYPE);
1931 range = (((s->out_format == FMT_MPEG1 || s->msmpeg4_version) ? 8 : 16) << f_code);
1933 assert(range <= 16 || !s->msmpeg4_version);
1934 assert(range <=256 || !(s->codec_id == CODEC_ID_MPEG2VIDEO && s->avctx->strict_std_compliance >= FF_COMPLIANCE_NORMAL));
1936 if(c->avctx->me_range && range > c->avctx->me_range) range= c->avctx->me_range;
1938 //printf("%d no:%d %d//\n", clip, noclip, f_code);
1939 if(s->flags&CODEC_FLAG_4MV){
1940 const int wrap= s->b8_stride;
1942 /* clip / convert to intra 8x8 type MVs */
1943 for(y=0; y<s->mb_height; y++){
1945 int i= y*s->mb_stride;
1948 for(x=0; x<s->mb_width; x++){
1949 if(s->mb_type[i]&CANDIDATE_MB_TYPE_INTER4V){
1951 for(block=0; block<4; block++){
1952 int off= (block& 1) + (block>>1)*wrap;
1953 int mx= s->current_picture.motion_val[0][ xy + off ][0];
1954 int my= s->current_picture.motion_val[0][ xy + off ][1];
1956 if( mx >=range || mx <-range
1957 || my >=range || my <-range){
1958 s->mb_type[i] &= ~CANDIDATE_MB_TYPE_INTER4V;
1959 s->mb_type[i] |= CANDIDATE_MB_TYPE_INTRA;
1960 s->current_picture.mb_type[i]= CANDIDATE_MB_TYPE_INTRA;
1973 * @param truncate 1 for truncation, 0 for using intra
1975 void ff_fix_long_mvs(MpegEncContext * s, uint8_t *field_select_table, int field_select,
1976 int16_t (*mv_table)[2], int f_code, int type, int truncate)
1978 MotionEstContext * const c= &s->me;
1979 int y, h_range, v_range;
1981 // RAL: 8 in MPEG-1, 16 in MPEG-4
1982 int range = (((s->out_format == FMT_MPEG1 || s->msmpeg4_version) ? 8 : 16) << f_code);
1984 if(c->avctx->me_range && range > c->avctx->me_range) range= c->avctx->me_range;
1987 v_range= field_select_table ? range>>1 : range;
1989 /* clip / convert to intra 16x16 type MVs */
1990 for(y=0; y<s->mb_height; y++){
1992 int xy= y*s->mb_stride;
1993 for(x=0; x<s->mb_width; x++){
1994 if (s->mb_type[xy] & type){ // RAL: "type" test added...
1995 if(field_select_table==NULL || field_select_table[xy] == field_select){
1996 if( mv_table[xy][0] >=h_range || mv_table[xy][0] <-h_range
1997 || mv_table[xy][1] >=v_range || mv_table[xy][1] <-v_range){
2000 if (mv_table[xy][0] > h_range-1) mv_table[xy][0]= h_range-1;
2001 else if(mv_table[xy][0] < -h_range ) mv_table[xy][0]= -h_range;
2002 if (mv_table[xy][1] > v_range-1) mv_table[xy][1]= v_range-1;
2003 else if(mv_table[xy][1] < -v_range ) mv_table[xy][1]= -v_range;
2005 s->mb_type[xy] &= ~type;
2006 s->mb_type[xy] |= CANDIDATE_MB_TYPE_INTRA;