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
36 #include "mpegutils.h"
37 #include "mpegvideo.h"
44 #define P_TOPRIGHT P[3]
48 #define ME_MAP_SHIFT 3
49 #define ME_MAP_MV_BITS 11
51 static int sad_hpel_motion_search(MpegEncContext * s,
52 int *mx_ptr, int *my_ptr, int dmin,
53 int src_index, int ref_index,
56 static inline unsigned update_map_generation(MotionEstContext *c)
58 c->map_generation+= 1<<(ME_MAP_MV_BITS*2);
59 if(c->map_generation==0){
60 c->map_generation= 1<<(ME_MAP_MV_BITS*2);
61 memset(c->map, 0, sizeof(uint32_t)*ME_MAP_SIZE);
63 return c->map_generation;
66 /* shape adaptive search stuff */
67 typedef struct Minima{
73 static int minima_cmp(const void *a, const void *b){
74 const Minima *da = (const Minima *) a;
75 const Minima *db = (const Minima *) b;
77 return da->height - db->height;
80 #define FLAG_QPEL 1 //must be 1
84 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){
85 const int offset[3]= {
87 ((y*c->uvstride + x)>>1),
88 ((y*c->uvstride + x)>>1),
92 c->src[0][i]= src [i] + offset[i];
93 c->ref[0][i]= ref [i] + offset[i];
97 c->ref[ref_index][i]= ref2[i] + offset[i];
102 static int get_flags(MotionEstContext *c, int direct, int chroma){
103 return ((c->avctx->flags&CODEC_FLAG_QPEL) ? FLAG_QPEL : 0)
104 + (direct ? FLAG_DIRECT : 0)
105 + (chroma ? FLAG_CHROMA : 0);
108 static av_always_inline int cmp_direct_inline(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
109 const int size, const int h, int ref_index, int src_index,
110 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, int qpel){
111 MotionEstContext * const c= &s->me;
112 const int stride= c->stride;
113 const int hx= subx + (x<<(1+qpel));
114 const int hy= suby + (y<<(1+qpel));
115 uint8_t * const * const ref= c->ref[ref_index];
116 uint8_t * const * const src= c->src[src_index];
118 //FIXME check chroma 4mv, (no crashes ...)
119 av_assert2(x >= c->xmin && hx <= c->xmax<<(qpel+1) && y >= c->ymin && hy <= c->ymax<<(qpel+1));
120 if(x >= c->xmin && hx <= c->xmax<<(qpel+1) && y >= c->ymin && hy <= c->ymax<<(qpel+1)){
121 const int time_pp= s->pp_time;
122 const int time_pb= s->pb_time;
123 const int mask= 2*qpel+1;
124 if(s->mv_type==MV_TYPE_8X8){
127 int fx = c->direct_basis_mv[i][0] + hx;
128 int fy = c->direct_basis_mv[i][1] + hy;
129 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));
130 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));
131 int fxy= (fx&mask) + ((fy&mask)<<(qpel+1));
132 int bxy= (bx&mask) + ((by&mask)<<(qpel+1));
134 uint8_t *dst= c->temp + 8*(i&1) + 8*stride*(i>>1);
136 c->qpel_put[1][fxy](dst, ref[0] + (fx>>2) + (fy>>2)*stride, stride);
137 c->qpel_avg[1][bxy](dst, ref[8] + (bx>>2) + (by>>2)*stride, stride);
139 c->hpel_put[1][fxy](dst, ref[0] + (fx>>1) + (fy>>1)*stride, stride, 8);
140 c->hpel_avg[1][bxy](dst, ref[8] + (bx>>1) + (by>>1)*stride, stride, 8);
144 int fx = c->direct_basis_mv[0][0] + hx;
145 int fy = c->direct_basis_mv[0][1] + hy;
146 int bx = hx ? fx - c->co_located_mv[0][0] : (c->co_located_mv[0][0]*(time_pb - time_pp)/time_pp);
147 int by = hy ? fy - c->co_located_mv[0][1] : (c->co_located_mv[0][1]*(time_pb - time_pp)/time_pp);
148 int fxy= (fx&mask) + ((fy&mask)<<(qpel+1));
149 int bxy= (bx&mask) + ((by&mask)<<(qpel+1));
152 c->qpel_put[1][fxy](c->temp , ref[0] + (fx>>2) + (fy>>2)*stride , stride);
153 c->qpel_put[1][fxy](c->temp + 8 , ref[0] + (fx>>2) + (fy>>2)*stride + 8 , stride);
154 c->qpel_put[1][fxy](c->temp + 8*stride, ref[0] + (fx>>2) + (fy>>2)*stride + 8*stride, stride);
155 c->qpel_put[1][fxy](c->temp + 8 + 8*stride, ref[0] + (fx>>2) + (fy>>2)*stride + 8 + 8*stride, stride);
156 c->qpel_avg[1][bxy](c->temp , ref[8] + (bx>>2) + (by>>2)*stride , stride);
157 c->qpel_avg[1][bxy](c->temp + 8 , ref[8] + (bx>>2) + (by>>2)*stride + 8 , stride);
158 c->qpel_avg[1][bxy](c->temp + 8*stride, ref[8] + (bx>>2) + (by>>2)*stride + 8*stride, stride);
159 c->qpel_avg[1][bxy](c->temp + 8 + 8*stride, ref[8] + (bx>>2) + (by>>2)*stride + 8 + 8*stride, stride);
161 av_assert2((fx>>1) + 16*s->mb_x >= -16);
162 av_assert2((fy>>1) + 16*s->mb_y >= -16);
163 av_assert2((fx>>1) + 16*s->mb_x <= s->width);
164 av_assert2((fy>>1) + 16*s->mb_y <= s->height);
165 av_assert2((bx>>1) + 16*s->mb_x >= -16);
166 av_assert2((by>>1) + 16*s->mb_y >= -16);
167 av_assert2((bx>>1) + 16*s->mb_x <= s->width);
168 av_assert2((by>>1) + 16*s->mb_y <= s->height);
170 c->hpel_put[0][fxy](c->temp, ref[0] + (fx>>1) + (fy>>1)*stride, stride, 16);
171 c->hpel_avg[0][bxy](c->temp, ref[8] + (bx>>1) + (by>>1)*stride, stride, 16);
174 d = cmp_func(s, c->temp, src[0], stride, 16);
180 static av_always_inline int cmp_inline(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
181 const int size, const int h, int ref_index, int src_index,
182 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, int qpel, int chroma){
183 MotionEstContext * const c= &s->me;
184 const int stride= c->stride;
185 const int uvstride= c->uvstride;
186 const int dxy= subx + (suby<<(1+qpel)); //FIXME log2_subpel?
187 const int hx= subx + (x<<(1+qpel));
188 const int hy= suby + (y<<(1+qpel));
189 uint8_t * const * const ref= c->ref[ref_index];
190 uint8_t * const * const src= c->src[src_index];
192 //FIXME check chroma 4mv, (no crashes ...)
193 int uvdxy; /* no, it might not be used uninitialized */
196 c->qpel_put[size][dxy](c->temp, ref[0] + x + y*stride, stride); //FIXME prototype (add h)
202 uvdxy= (cx&1) + 2*(cy&1);
203 //FIXME x/y wrong, but mpeg4 qpel is sick anyway, we should drop as much of it as possible in favor for h264
206 c->hpel_put[size][dxy](c->temp, ref[0] + x + y*stride, stride, h);
208 uvdxy= dxy | (x&1) | (2*(y&1));
210 d = cmp_func(s, c->temp, src[0], stride, h);
212 d = cmp_func(s, src[0], ref[0] + x + y*stride, stride, h);
214 uvdxy= (x&1) + 2*(y&1);
217 uint8_t * const uvtemp= c->temp + 16*stride;
218 c->hpel_put[size+1][uvdxy](uvtemp , ref[1] + (x>>1) + (y>>1)*uvstride, uvstride, h>>1);
219 c->hpel_put[size+1][uvdxy](uvtemp+8, ref[2] + (x>>1) + (y>>1)*uvstride, uvstride, h>>1);
220 d += chroma_cmp_func(s, uvtemp , src[1], uvstride, h>>1);
221 d += chroma_cmp_func(s, uvtemp+8, src[2], uvstride, h>>1);
226 static int cmp_simple(MpegEncContext *s, const int x, const int y,
227 int ref_index, int src_index,
228 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func){
229 return cmp_inline(s,x,y,0,0,0,16,ref_index,src_index, cmp_func, chroma_cmp_func, 0, 0);
232 static int cmp_fpel_internal(MpegEncContext *s, const int x, const int y,
233 const int size, const int h, int ref_index, int src_index,
234 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
235 if(flags&FLAG_DIRECT){
236 return cmp_direct_inline(s,x,y,0,0,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, flags&FLAG_QPEL);
238 return cmp_inline(s,x,y,0,0,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 0, flags&FLAG_CHROMA);
242 static int cmp_internal(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
243 const int size, const int h, int ref_index, int src_index,
244 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
245 if(flags&FLAG_DIRECT){
246 return cmp_direct_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, flags&FLAG_QPEL);
248 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);
252 /** @brief compares a block (either a full macroblock or a partition thereof)
253 against a proposed motion-compensated prediction of that block
255 static av_always_inline int cmp(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
256 const int size, const int h, int ref_index, int src_index,
257 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
258 if(av_builtin_constant_p(flags) && av_builtin_constant_p(h) && av_builtin_constant_p(size)
259 && av_builtin_constant_p(subx) && av_builtin_constant_p(suby)
260 && flags==0 && h==16 && size==0 && subx==0 && suby==0){
261 return cmp_simple(s,x,y,ref_index,src_index, cmp_func, chroma_cmp_func);
262 }else if(av_builtin_constant_p(subx) && av_builtin_constant_p(suby)
263 && subx==0 && suby==0){
264 return cmp_fpel_internal(s,x,y,size,h,ref_index,src_index, cmp_func, chroma_cmp_func,flags);
266 return cmp_internal(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, flags);
270 static int cmp_hpel(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
271 const int size, const int h, int ref_index, int src_index,
272 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
273 if(flags&FLAG_DIRECT){
274 return cmp_direct_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 0);
276 return cmp_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 0, flags&FLAG_CHROMA);
280 static int cmp_qpel(MpegEncContext *s, const int x, const int y, const int subx, const int suby,
281 const int size, const int h, int ref_index, int src_index,
282 me_cmp_func cmp_func, me_cmp_func chroma_cmp_func, const int flags){
283 if(flags&FLAG_DIRECT){
284 return cmp_direct_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 1);
286 return cmp_inline(s,x,y,subx,suby,size,h,ref_index,src_index, cmp_func, chroma_cmp_func, 1, flags&FLAG_CHROMA);
290 #include "motion_est_template.c"
292 static int zero_cmp(MpegEncContext *s, uint8_t *a, uint8_t *b,
298 static void zero_hpel(uint8_t *a, const uint8_t *b, ptrdiff_t stride, int h){
301 int ff_init_me(MpegEncContext *s){
302 MotionEstContext * const c= &s->me;
303 int cache_size= FFMIN(ME_MAP_SIZE>>ME_MAP_SHIFT, 1<<ME_MAP_SHIFT);
304 int dia_size= FFMAX(FFABS(s->avctx->dia_size)&255, FFABS(s->avctx->pre_dia_size)&255);
306 if(FFMIN(s->avctx->dia_size, s->avctx->pre_dia_size) < -FFMIN(ME_MAP_SIZE, MAX_SAB_SIZE)){
307 av_log(s->avctx, AV_LOG_ERROR, "ME_MAP size is too small for SAB diamond\n");
310 //special case of snow is needed because snow uses its own iterative ME code
311 if(s->me_method!=ME_ZERO && s->me_method!=ME_EPZS && s->me_method!=ME_X1 && s->avctx->codec_id != AV_CODEC_ID_SNOW){
312 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");
318 if(cache_size < 2*dia_size && !c->stride){
319 av_log(s->avctx, AV_LOG_INFO, "ME_MAP size may be a little small for the selected diamond size\n");
322 ff_set_cmp(&s->mecc, s->mecc.me_pre_cmp, c->avctx->me_pre_cmp);
323 ff_set_cmp(&s->mecc, s->mecc.me_cmp, c->avctx->me_cmp);
324 ff_set_cmp(&s->mecc, s->mecc.me_sub_cmp, c->avctx->me_sub_cmp);
325 ff_set_cmp(&s->mecc, s->mecc.mb_cmp, c->avctx->mb_cmp);
327 c->flags = get_flags(c, 0, c->avctx->me_cmp &FF_CMP_CHROMA);
328 c->sub_flags= get_flags(c, 0, c->avctx->me_sub_cmp&FF_CMP_CHROMA);
329 c->mb_flags = get_flags(c, 0, c->avctx->mb_cmp &FF_CMP_CHROMA);
331 /*FIXME s->no_rounding b_type*/
332 if(s->flags&CODEC_FLAG_QPEL){
333 c->sub_motion_search= qpel_motion_search;
334 c->qpel_avg = s->qdsp.avg_qpel_pixels_tab;
336 c->qpel_put = s->qdsp.put_no_rnd_qpel_pixels_tab;
338 c->qpel_put = s->qdsp.put_qpel_pixels_tab;
340 if(c->avctx->me_sub_cmp&FF_CMP_CHROMA)
341 c->sub_motion_search= hpel_motion_search;
342 else if( c->avctx->me_sub_cmp == FF_CMP_SAD
343 && c->avctx-> me_cmp == FF_CMP_SAD
344 && c->avctx-> mb_cmp == FF_CMP_SAD)
345 c->sub_motion_search= sad_hpel_motion_search; // 2050 vs. 2450 cycles
347 c->sub_motion_search= hpel_motion_search;
349 c->hpel_avg = s->hdsp.avg_pixels_tab;
351 c->hpel_put = s->hdsp.put_no_rnd_pixels_tab;
353 c->hpel_put = s->hdsp.put_pixels_tab;
356 c->stride = s->linesize;
357 c->uvstride= s->uvlinesize;
359 c->stride = 16*s->mb_width + 32;
360 c->uvstride= 8*s->mb_width + 16;
363 /* 8x8 fullpel search would need a 4x4 chroma compare, which we do
364 * not have yet, and even if we had, the motion estimation code
365 * does not expect it. */
366 if (s->codec_id != AV_CODEC_ID_SNOW) {
367 if ((c->avctx->me_cmp & FF_CMP_CHROMA) /* && !s->mecc.me_cmp[2] */)
368 s->mecc.me_cmp[2] = zero_cmp;
369 if ((c->avctx->me_sub_cmp & FF_CMP_CHROMA) && !s->mecc.me_sub_cmp[2])
370 s->mecc.me_sub_cmp[2] = zero_cmp;
371 c->hpel_put[2][0]= c->hpel_put[2][1]=
372 c->hpel_put[2][2]= c->hpel_put[2][3]= zero_hpel;
375 if(s->codec_id == AV_CODEC_ID_H261){
376 c->sub_motion_search= no_sub_motion_search;
382 #define CHECK_SAD_HALF_MV(suffix, x, y) \
384 d = s->mecc.pix_abs[size][(x ? 1 : 0) + (y ? 2 : 0)](NULL, pix, ptr + ((x) >> 1), stride, h); \
385 d += (mv_penalty[pen_x + x] + mv_penalty[pen_y + y])*penalty_factor;\
386 COPY3_IF_LT(dminh, d, dx, x, dy, y)\
389 static int sad_hpel_motion_search(MpegEncContext * s,
390 int *mx_ptr, int *my_ptr, int dmin,
391 int src_index, int ref_index,
394 MotionEstContext * const c= &s->me;
395 const int penalty_factor= c->sub_penalty_factor;
398 int stride= c->stride;
401 av_assert2(c->sub_flags == 0);
409 pix = c->src[src_index][0];
413 ptr = c->ref[ref_index][0] + (my * stride) + mx;
417 if (mx > xmin && mx < xmax &&
418 my > ymin && my < ymax) {
421 const int index= (my<<ME_MAP_SHIFT) + mx;
422 const int t= score_map[(index-(1<<ME_MAP_SHIFT))&(ME_MAP_SIZE-1)];
423 const int l= score_map[(index- 1 )&(ME_MAP_SIZE-1)];
424 const int r= score_map[(index+ 1 )&(ME_MAP_SIZE-1)];
425 const int b= score_map[(index+(1<<ME_MAP_SHIFT))&(ME_MAP_SIZE-1)];
435 CHECK_SAD_HALF_MV(y2 , 0, -1)
437 CHECK_SAD_HALF_MV(xy2, -1, -1)
439 CHECK_SAD_HALF_MV(xy2, +1, -1)
443 CHECK_SAD_HALF_MV(xy2, -1, +1)
445 CHECK_SAD_HALF_MV(x2 , -1, 0)
447 CHECK_SAD_HALF_MV(xy2, +1, -1)
449 CHECK_SAD_HALF_MV(xy2, -1, -1)
453 CHECK_SAD_HALF_MV(xy2, +1, +1)
455 CHECK_SAD_HALF_MV(x2 , +1, 0)
460 CHECK_SAD_HALF_MV(xy2, -1, -1)
464 CHECK_SAD_HALF_MV(xy2, +1, +1)
466 CHECK_SAD_HALF_MV(x2 , -1, 0)
467 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)
477 CHECK_SAD_HALF_MV(xy2, +1, +1)
479 CHECK_SAD_HALF_MV(y2 , 0, +1)
494 static inline void set_p_mv_tables(MpegEncContext * s, int mx, int my, int mv4)
496 const int xy= s->mb_x + s->mb_y*s->mb_stride;
498 s->p_mv_table[xy][0] = mx;
499 s->p_mv_table[xy][1] = my;
501 /* has already been set to the 4 MV if 4MV is done */
503 int mot_xy= s->block_index[0];
505 s->current_picture.motion_val[0][mot_xy ][0] = mx;
506 s->current_picture.motion_val[0][mot_xy ][1] = my;
507 s->current_picture.motion_val[0][mot_xy + 1][0] = mx;
508 s->current_picture.motion_val[0][mot_xy + 1][1] = my;
510 mot_xy += s->b8_stride;
511 s->current_picture.motion_val[0][mot_xy ][0] = mx;
512 s->current_picture.motion_val[0][mot_xy ][1] = my;
513 s->current_picture.motion_val[0][mot_xy + 1][0] = mx;
514 s->current_picture.motion_val[0][mot_xy + 1][1] = my;
519 * get fullpel ME search limits.
521 static inline void get_limits(MpegEncContext *s, int x, int y)
523 MotionEstContext * const c= &s->me;
524 int range= c->avctx->me_range >> (1 + !!(c->flags&FLAG_QPEL));
525 int max_range = MAX_MV >> (1 + !!(c->flags&FLAG_QPEL));
527 if(c->avctx->me_range) c->range= c->avctx->me_range >> 1;
530 if (s->unrestricted_mv) {
533 c->xmax = - x + s->width;
534 c->ymax = - y + s->height;
535 } else if (s->out_format == FMT_H261){
536 // Search range of H261 is different from other codec standards
537 c->xmin = (x > 15) ? - 15 : 0;
538 c->ymin = (y > 15) ? - 15 : 0;
539 c->xmax = (x < s->mb_width * 16 - 16) ? 15 : 0;
540 c->ymax = (y < s->mb_height * 16 - 16) ? 15 : 0;
544 c->xmax = - x + s->mb_width *16 - 16;
545 c->ymax = - y + s->mb_height*16 - 16;
547 if(!range || range > max_range)
550 c->xmin = FFMAX(c->xmin,-range);
551 c->xmax = FFMIN(c->xmax, range);
552 c->ymin = FFMAX(c->ymin,-range);
553 c->ymax = FFMIN(c->ymax, range);
557 static inline void init_mv4_ref(MotionEstContext *c){
558 const int stride= c->stride;
560 c->ref[1][0] = c->ref[0][0] + 8;
561 c->ref[2][0] = c->ref[0][0] + 8*stride;
562 c->ref[3][0] = c->ref[2][0] + 8;
563 c->src[1][0] = c->src[0][0] + 8;
564 c->src[2][0] = c->src[0][0] + 8*stride;
565 c->src[3][0] = c->src[2][0] + 8;
568 static inline int h263_mv4_search(MpegEncContext *s, int mx, int my, int shift)
570 MotionEstContext * const c= &s->me;
575 int dmin_sum=0, mx4_sum=0, my4_sum=0, i;
577 const int stride= c->stride;
578 uint8_t *mv_penalty= c->current_mv_penalty;
579 int saftey_cliping= s->unrestricted_mv && (s->width&15) && (s->height&15);
583 for(block=0; block<4; block++){
585 int pred_x4, pred_y4;
587 static const int off[4]= {2, 1, 1, -1};
588 const int mot_stride = s->b8_stride;
589 const int mot_xy = s->block_index[block];
592 c->xmax = - 16*s->mb_x + s->width - 8*(block &1);
593 c->ymax = - 16*s->mb_y + s->height - 8*(block>>1);
596 P_LEFT[0] = s->current_picture.motion_val[0][mot_xy - 1][0];
597 P_LEFT[1] = s->current_picture.motion_val[0][mot_xy - 1][1];
599 if(P_LEFT[0] > (c->xmax<<shift)) P_LEFT[0] = (c->xmax<<shift);
601 /* special case for first line */
602 if (s->first_slice_line && block<2) {
603 c->pred_x= pred_x4= P_LEFT[0];
604 c->pred_y= pred_y4= P_LEFT[1];
606 P_TOP[0] = s->current_picture.motion_val[0][mot_xy - mot_stride ][0];
607 P_TOP[1] = s->current_picture.motion_val[0][mot_xy - mot_stride ][1];
608 P_TOPRIGHT[0] = s->current_picture.motion_val[0][mot_xy - mot_stride + off[block]][0];
609 P_TOPRIGHT[1] = s->current_picture.motion_val[0][mot_xy - mot_stride + off[block]][1];
610 if(P_TOP[1] > (c->ymax<<shift)) P_TOP[1] = (c->ymax<<shift);
611 if(P_TOPRIGHT[0] < (c->xmin<<shift)) P_TOPRIGHT[0]= (c->xmin<<shift);
612 if(P_TOPRIGHT[0] > (c->xmax<<shift)) P_TOPRIGHT[0]= (c->xmax<<shift);
613 if(P_TOPRIGHT[1] > (c->ymax<<shift)) P_TOPRIGHT[1]= (c->ymax<<shift);
615 P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
616 P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
618 c->pred_x= pred_x4 = P_MEDIAN[0];
619 c->pred_y= pred_y4 = P_MEDIAN[1];
625 if (s->first_slice_line && block<2 && i>1 && i<9)
629 if(P[i][0] > (c->xmax<<shift)) P[i][0]= (c->xmax<<shift);
630 if(P[i][1] > (c->ymax<<shift)) P[i][1]= (c->ymax<<shift);
633 dmin4 = epzs_motion_search4(s, &mx4, &my4, P, block, block, s->p_mv_table, (1<<16)>>shift);
635 dmin4= c->sub_motion_search(s, &mx4, &my4, dmin4, block, block, size, h);
637 if (s->mecc.me_sub_cmp[0] != s->mecc.mb_cmp[0]) {
639 const int offset= ((block&1) + (block>>1)*stride)*8;
640 uint8_t *dest_y = c->scratchpad + offset;
641 if(s->quarter_sample){
642 uint8_t *ref= c->ref[block][0] + (mx4>>2) + (my4>>2)*stride;
643 dxy = ((my4 & 3) << 2) | (mx4 & 3);
646 s->qdsp.put_no_rnd_qpel_pixels_tab[1][dxy](dest_y, ref, stride);
648 s->qdsp.put_qpel_pixels_tab[1][dxy](dest_y, ref, stride);
650 uint8_t *ref= c->ref[block][0] + (mx4>>1) + (my4>>1)*stride;
651 dxy = ((my4 & 1) << 1) | (mx4 & 1);
654 s->hdsp.put_no_rnd_pixels_tab[1][dxy](dest_y , ref , stride, h);
656 s->hdsp.put_pixels_tab [1][dxy](dest_y , ref , stride, h);
658 dmin_sum+= (mv_penalty[mx4-pred_x4] + mv_penalty[my4-pred_y4])*c->mb_penalty_factor;
662 if(s->quarter_sample){
670 s->current_picture.motion_val[0][s->block_index[block]][0] = mx4;
671 s->current_picture.motion_val[0][s->block_index[block]][1] = my4;
673 if(mx4 != mx || my4 != my) same=0;
679 if (s->mecc.me_sub_cmp[0] != s->mecc.mb_cmp[0]) {
680 dmin_sum += s->mecc.mb_cmp[0](s,
681 s->new_picture.f->data[0] +
682 s->mb_x * 16 + s->mb_y * 16 * stride,
683 c->scratchpad, stride, 16);
686 if(c->avctx->mb_cmp&FF_CMP_CHROMA){
691 mx= ff_h263_round_chroma(mx4_sum);
692 my= ff_h263_round_chroma(my4_sum);
693 dxy = ((my & 1) << 1) | (mx & 1);
695 offset= (s->mb_x*8 + (mx>>1)) + (s->mb_y*8 + (my>>1))*s->uvlinesize;
698 s->hdsp.put_no_rnd_pixels_tab[1][dxy](c->scratchpad , s->last_picture.f->data[1] + offset, s->uvlinesize, 8);
699 s->hdsp.put_no_rnd_pixels_tab[1][dxy](c->scratchpad + 8, s->last_picture.f->data[2] + offset, s->uvlinesize, 8);
701 s->hdsp.put_pixels_tab [1][dxy](c->scratchpad , s->last_picture.f->data[1] + offset, s->uvlinesize, 8);
702 s->hdsp.put_pixels_tab [1][dxy](c->scratchpad + 8, s->last_picture.f->data[2] + offset, s->uvlinesize, 8);
705 dmin_sum += s->mecc.mb_cmp[1](s, s->new_picture.f->data[1] + s->mb_x * 8 + s->mb_y * 8 * s->uvlinesize, c->scratchpad, s->uvlinesize, 8);
706 dmin_sum += s->mecc.mb_cmp[1](s, s->new_picture.f->data[2] + s->mb_x * 8 + s->mb_y * 8 * s->uvlinesize, c->scratchpad + 8, s->uvlinesize, 8);
712 switch(c->avctx->mb_cmp&0xFF){
714 return dmin_sum+ 32*s->qscale*s->qscale;*/
718 return dmin_sum+ 11*c->mb_penalty_factor;
722 static inline void init_interlaced_ref(MpegEncContext *s, int ref_index){
723 MotionEstContext * const c= &s->me;
725 c->ref[1+ref_index][0] = c->ref[0+ref_index][0] + s->linesize;
726 c->src[1][0] = c->src[0][0] + s->linesize;
727 if(c->flags & FLAG_CHROMA){
728 c->ref[1+ref_index][1] = c->ref[0+ref_index][1] + s->uvlinesize;
729 c->ref[1+ref_index][2] = c->ref[0+ref_index][2] + s->uvlinesize;
730 c->src[1][1] = c->src[0][1] + s->uvlinesize;
731 c->src[1][2] = c->src[0][2] + s->uvlinesize;
735 static int interlaced_search(MpegEncContext *s, int ref_index,
736 int16_t (*mv_tables[2][2])[2], uint8_t *field_select_tables[2], int mx, int my, int user_field_select)
738 MotionEstContext * const c= &s->me;
743 uint8_t * const mv_penalty= c->current_mv_penalty;
745 const int stride= 2*s->linesize;
747 const int mot_stride= s->mb_stride;
748 const int xy= s->mb_x + s->mb_y*mot_stride;
754 init_interlaced_ref(s, ref_index);
756 for(block=0; block<2; block++){
758 int best_dmin= INT_MAX;
761 for(field_select=0; field_select<2; field_select++){
762 int dmin, mx_i, my_i;
763 int16_t (*mv_table)[2]= mv_tables[block][field_select];
765 if(user_field_select){
766 av_assert1(field_select==0 || field_select==1);
767 av_assert1(field_select_tables[block][xy]==0 || field_select_tables[block][xy]==1);
768 if(field_select_tables[block][xy] != field_select)
772 P_LEFT[0] = mv_table[xy - 1][0];
773 P_LEFT[1] = mv_table[xy - 1][1];
774 if(P_LEFT[0] > (c->xmax<<1)) P_LEFT[0] = (c->xmax<<1);
776 c->pred_x= P_LEFT[0];
777 c->pred_y= P_LEFT[1];
779 if(!s->first_slice_line){
780 P_TOP[0] = mv_table[xy - mot_stride][0];
781 P_TOP[1] = mv_table[xy - mot_stride][1];
782 P_TOPRIGHT[0] = mv_table[xy - mot_stride + 1][0];
783 P_TOPRIGHT[1] = mv_table[xy - mot_stride + 1][1];
784 if(P_TOP[1] > (c->ymax<<1)) P_TOP[1] = (c->ymax<<1);
785 if(P_TOPRIGHT[0] < (c->xmin<<1)) P_TOPRIGHT[0]= (c->xmin<<1);
786 if(P_TOPRIGHT[0] > (c->xmax<<1)) P_TOPRIGHT[0]= (c->xmax<<1);
787 if(P_TOPRIGHT[1] > (c->ymax<<1)) P_TOPRIGHT[1]= (c->ymax<<1);
789 P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
790 P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
792 P_MV1[0]= mx; //FIXME not correct if block != field_select
795 dmin = epzs_motion_search2(s, &mx_i, &my_i, P, block, field_select+ref_index, mv_table, (1<<16)>>1);
797 dmin= c->sub_motion_search(s, &mx_i, &my_i, dmin, block, field_select+ref_index, size, h);
799 mv_table[xy][0]= mx_i;
800 mv_table[xy][1]= my_i;
802 if (s->mecc.me_sub_cmp[0] != s->mecc.mb_cmp[0]) {
806 uint8_t *ref= c->ref[field_select+ref_index][0] + (mx_i>>1) + (my_i>>1)*stride;
807 dxy = ((my_i & 1) << 1) | (mx_i & 1);
810 s->hdsp.put_no_rnd_pixels_tab[size][dxy](c->scratchpad, ref , stride, h);
812 s->hdsp.put_pixels_tab [size][dxy](c->scratchpad, ref , stride, h);
814 dmin = s->mecc.mb_cmp[size](s, c->src[block][0], c->scratchpad, stride, h);
815 dmin+= (mv_penalty[mx_i-c->pred_x] + mv_penalty[my_i-c->pred_y] + 1)*c->mb_penalty_factor;
817 dmin+= c->mb_penalty_factor; //field_select bits
819 dmin += field_select != block; //slightly prefer same field
821 if(dmin < best_dmin){
823 best_field= field_select;
827 int16_t (*mv_table)[2]= mv_tables[block][best_field];
829 if(mv_table[xy][0] != mx) same=0; //FIXME check if these checks work and are any good at all
830 if(mv_table[xy][1]&1) same=0;
831 if(mv_table[xy][1]*2 != my) same=0;
832 if(best_field != block) same=0;
835 field_select_tables[block][xy]= best_field;
836 dmin_sum += best_dmin;
847 switch(c->avctx->mb_cmp&0xFF){
849 return dmin_sum+ 32*s->qscale*s->qscale;*/
853 return dmin_sum+ 11*c->mb_penalty_factor;
857 static inline int get_penalty_factor(int lambda, int lambda2, int type){
861 return lambda>>FF_LAMBDA_SHIFT;
863 return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
865 return (4*lambda)>>(FF_LAMBDA_SHIFT);
867 return (2*lambda)>>(FF_LAMBDA_SHIFT);
870 return (2*lambda)>>FF_LAMBDA_SHIFT;
875 return lambda2>>FF_LAMBDA_SHIFT;
881 void ff_estimate_p_frame_motion(MpegEncContext * s,
884 MotionEstContext * const c= &s->me;
886 int sum, mx, my, dmin;
887 int varc; ///< the variance of the block (sum of squared (p[y][x]-average))
888 int vard; ///< sum of squared differences with the estimated motion vector
890 const int shift= 1+s->quarter_sample;
892 Picture * const pic= &s->current_picture;
894 init_ref(c, s->new_picture.f->data, s->last_picture.f->data, NULL, 16*mb_x, 16*mb_y, 0);
896 av_assert0(s->quarter_sample==0 || s->quarter_sample==1);
897 av_assert0(s->linesize == c->stride);
898 av_assert0(s->uvlinesize == c->uvstride);
900 c->penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_cmp);
901 c->sub_penalty_factor= get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_sub_cmp);
902 c->mb_penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->mb_cmp);
903 c->current_mv_penalty= c->mv_penalty[s->f_code] + MAX_MV;
905 get_limits(s, 16*mb_x, 16*mb_y);
908 /* intra / predictive decision */
910 sum = s->mpvencdsp.pix_sum(pix, s->linesize);
911 varc = s->mpvencdsp.pix_norm1(pix, s->linesize) -
912 (((unsigned) sum * sum) >> 8) + 500;
914 pic->mb_mean[s->mb_stride * mb_y + mb_x] = (sum+128)>>8;
915 pic->mb_var [s->mb_stride * mb_y + mb_x] = (varc+128)>>8;
916 c->mb_var_sum_temp += (varc+128)>>8;
918 switch(s->me_method) {
928 const int mot_stride = s->b8_stride;
929 const int mot_xy = s->block_index[0];
931 P_LEFT[0] = s->current_picture.motion_val[0][mot_xy - 1][0];
932 P_LEFT[1] = s->current_picture.motion_val[0][mot_xy - 1][1];
934 if(P_LEFT[0] > (c->xmax<<shift)) P_LEFT[0] = (c->xmax<<shift);
936 if(!s->first_slice_line) {
937 P_TOP[0] = s->current_picture.motion_val[0][mot_xy - mot_stride ][0];
938 P_TOP[1] = s->current_picture.motion_val[0][mot_xy - mot_stride ][1];
939 P_TOPRIGHT[0] = s->current_picture.motion_val[0][mot_xy - mot_stride + 2][0];
940 P_TOPRIGHT[1] = s->current_picture.motion_val[0][mot_xy - mot_stride + 2][1];
941 if(P_TOP[1] > (c->ymax<<shift)) P_TOP[1] = (c->ymax<<shift);
942 if(P_TOPRIGHT[0] < (c->xmin<<shift)) P_TOPRIGHT[0]= (c->xmin<<shift);
943 if(P_TOPRIGHT[1] > (c->ymax<<shift)) P_TOPRIGHT[1]= (c->ymax<<shift);
945 P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
946 P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
948 if(s->out_format == FMT_H263){
949 c->pred_x = P_MEDIAN[0];
950 c->pred_y = P_MEDIAN[1];
951 }else { /* mpeg1 at least */
952 c->pred_x= P_LEFT[0];
953 c->pred_y= P_LEFT[1];
956 c->pred_x= P_LEFT[0];
957 c->pred_y= P_LEFT[1];
961 dmin = ff_epzs_motion_search(s, &mx, &my, P, 0, 0, s->p_mv_table, (1<<16)>>shift, 0, 16);
966 /* At this point (mx,my) are full-pell and the relative displacement */
967 ppix = c->ref[0][0] + (my * s->linesize) + mx;
969 vard = s->mecc.sse[0](NULL, pix, ppix, s->linesize, 16);
971 pic->mc_mb_var[s->mb_stride * mb_y + mb_x] = (vard+128)>>8;
972 c->mc_mb_var_sum_temp += (vard+128)>>8;
975 int p_score= FFMIN(vard, varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*100);
976 int i_score= varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*20;
977 c->scene_change_score+= ff_sqrt(p_score) - ff_sqrt(i_score);
979 if(mb_type == CANDIDATE_MB_TYPE_INTER){
980 c->sub_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
981 set_p_mv_tables(s, mx, my, 1);
986 if(mb_type == CANDIDATE_MB_TYPE_INTER4V){
987 h263_mv4_search(s, mx, my, shift);
989 set_p_mv_tables(s, mx, my, 0);
991 if(mb_type == CANDIDATE_MB_TYPE_INTER_I){
992 interlaced_search(s, 0, s->p_field_mv_table, s->p_field_select_table, mx, my, 1);
994 }else if(c->avctx->mb_decision > FF_MB_DECISION_SIMPLE){
995 int p_score= FFMIN(vard, varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*100);
996 int i_score= varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*20;
997 c->scene_change_score+= ff_sqrt(p_score) - ff_sqrt(i_score);
999 if (vard*2 + 200*256 > varc)
1000 mb_type|= CANDIDATE_MB_TYPE_INTRA;
1001 if (varc*2 + 200*256 > vard || s->qscale > 24){
1002 // if (varc*2 + 200*256 + 50*(s->lambda2>>FF_LAMBDA_SHIFT) > vard){
1003 mb_type|= CANDIDATE_MB_TYPE_INTER;
1004 c->sub_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
1005 if (s->mpv_flags & FF_MPV_FLAG_MV0)
1007 mb_type |= CANDIDATE_MB_TYPE_SKIPPED; //FIXME check difference
1012 if((s->flags&CODEC_FLAG_4MV)
1013 && !c->skip && varc>50<<8 && vard>10<<8){
1014 if(h263_mv4_search(s, mx, my, shift) < INT_MAX)
1015 mb_type|=CANDIDATE_MB_TYPE_INTER4V;
1017 set_p_mv_tables(s, mx, my, 0);
1019 set_p_mv_tables(s, mx, my, 1);
1020 if((s->flags&CODEC_FLAG_INTERLACED_ME)
1021 && !c->skip){ //FIXME varc/d checks
1022 if(interlaced_search(s, 0, s->p_field_mv_table, s->p_field_select_table, mx, my, 0) < INT_MAX)
1023 mb_type |= CANDIDATE_MB_TYPE_INTER_I;
1027 mb_type= CANDIDATE_MB_TYPE_INTER;
1029 dmin= c->sub_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
1030 if(c->avctx->me_sub_cmp != c->avctx->mb_cmp && !c->skip)
1031 dmin= get_mb_score(s, mx, my, 0, 0, 0, 16, 1);
1033 if((s->flags&CODEC_FLAG_4MV)
1034 && !c->skip && varc>50<<8 && vard>10<<8){
1035 int dmin4= h263_mv4_search(s, mx, my, shift);
1037 mb_type= CANDIDATE_MB_TYPE_INTER4V;
1041 if((s->flags&CODEC_FLAG_INTERLACED_ME)
1042 && !c->skip){ //FIXME varc/d checks
1043 int dmin_i= interlaced_search(s, 0, s->p_field_mv_table, s->p_field_select_table, mx, my, 0);
1045 mb_type = CANDIDATE_MB_TYPE_INTER_I;
1050 set_p_mv_tables(s, mx, my, mb_type!=CANDIDATE_MB_TYPE_INTER4V);
1052 /* get intra luma score */
1053 if((c->avctx->mb_cmp&0xFF)==FF_CMP_SSE){
1054 intra_score= varc - 500;
1056 unsigned mean = (sum+128)>>8;
1059 for(i=0; i<16; i++){
1060 *(uint32_t*)(&c->scratchpad[i*s->linesize+ 0]) = mean;
1061 *(uint32_t*)(&c->scratchpad[i*s->linesize+ 4]) = mean;
1062 *(uint32_t*)(&c->scratchpad[i*s->linesize+ 8]) = mean;
1063 *(uint32_t*)(&c->scratchpad[i*s->linesize+12]) = mean;
1066 intra_score= s->mecc.mb_cmp[0](s, c->scratchpad, pix, s->linesize, 16);
1068 intra_score += c->mb_penalty_factor*16;
1070 if(intra_score < dmin){
1071 mb_type= CANDIDATE_MB_TYPE_INTRA;
1072 s->current_picture.mb_type[mb_y*s->mb_stride + mb_x] = CANDIDATE_MB_TYPE_INTRA; //FIXME cleanup
1074 s->current_picture.mb_type[mb_y*s->mb_stride + mb_x] = 0;
1077 int p_score= FFMIN(vard, varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*100);
1078 int i_score= varc-500+(s->lambda2>>FF_LAMBDA_SHIFT)*20;
1079 c->scene_change_score+= ff_sqrt(p_score) - ff_sqrt(i_score);
1083 s->mb_type[mb_y*s->mb_stride + mb_x]= mb_type;
1086 int ff_pre_estimate_p_frame_motion(MpegEncContext * s,
1089 MotionEstContext * const c= &s->me;
1092 const int shift= 1+s->quarter_sample;
1093 const int xy= mb_x + mb_y*s->mb_stride;
1094 init_ref(c, s->new_picture.f->data, s->last_picture.f->data, NULL, 16*mb_x, 16*mb_y, 0);
1096 av_assert0(s->quarter_sample==0 || s->quarter_sample==1);
1098 c->pre_penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_pre_cmp);
1099 c->current_mv_penalty= c->mv_penalty[s->f_code] + MAX_MV;
1101 get_limits(s, 16*mb_x, 16*mb_y);
1104 P_LEFT[0] = s->p_mv_table[xy + 1][0];
1105 P_LEFT[1] = s->p_mv_table[xy + 1][1];
1107 if(P_LEFT[0] < (c->xmin<<shift)) P_LEFT[0] = (c->xmin<<shift);
1109 /* special case for first line */
1110 if (s->first_slice_line) {
1111 c->pred_x= P_LEFT[0];
1112 c->pred_y= P_LEFT[1];
1113 P_TOP[0]= P_TOPRIGHT[0]= P_MEDIAN[0]=
1114 P_TOP[1]= P_TOPRIGHT[1]= P_MEDIAN[1]= 0; //FIXME
1116 P_TOP[0] = s->p_mv_table[xy + s->mb_stride ][0];
1117 P_TOP[1] = s->p_mv_table[xy + s->mb_stride ][1];
1118 P_TOPRIGHT[0] = s->p_mv_table[xy + s->mb_stride - 1][0];
1119 P_TOPRIGHT[1] = s->p_mv_table[xy + s->mb_stride - 1][1];
1120 if(P_TOP[1] < (c->ymin<<shift)) P_TOP[1] = (c->ymin<<shift);
1121 if(P_TOPRIGHT[0] > (c->xmax<<shift)) P_TOPRIGHT[0]= (c->xmax<<shift);
1122 if(P_TOPRIGHT[1] < (c->ymin<<shift)) P_TOPRIGHT[1]= (c->ymin<<shift);
1124 P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
1125 P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
1127 c->pred_x = P_MEDIAN[0];
1128 c->pred_y = P_MEDIAN[1];
1131 dmin = ff_epzs_motion_search(s, &mx, &my, P, 0, 0, s->p_mv_table, (1<<16)>>shift, 0, 16);
1133 s->p_mv_table[xy][0] = mx<<shift;
1134 s->p_mv_table[xy][1] = my<<shift;
1139 static int estimate_motion_b(MpegEncContext *s, int mb_x, int mb_y,
1140 int16_t (*mv_table)[2], int ref_index, int f_code)
1142 MotionEstContext * const c= &s->me;
1145 const int shift= 1+s->quarter_sample;
1146 const int mot_stride = s->mb_stride;
1147 const int mot_xy = mb_y*mot_stride + mb_x;
1148 uint8_t * const mv_penalty= c->mv_penalty[f_code] + MAX_MV;
1151 c->penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_cmp);
1152 c->sub_penalty_factor= get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_sub_cmp);
1153 c->mb_penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->mb_cmp);
1154 c->current_mv_penalty= mv_penalty;
1156 get_limits(s, 16*mb_x, 16*mb_y);
1158 switch(s->me_method) {
1167 P_LEFT[0] = mv_table[mot_xy - 1][0];
1168 P_LEFT[1] = mv_table[mot_xy - 1][1];
1170 if (P_LEFT[0] > (c->xmax << shift)) P_LEFT[0] = (c->xmax << shift);
1172 /* special case for first line */
1173 if (!s->first_slice_line) {
1174 P_TOP[0] = mv_table[mot_xy - mot_stride ][0];
1175 P_TOP[1] = mv_table[mot_xy - mot_stride ][1];
1176 P_TOPRIGHT[0] = mv_table[mot_xy - mot_stride + 1][0];
1177 P_TOPRIGHT[1] = mv_table[mot_xy - mot_stride + 1][1];
1178 if (P_TOP[1] > (c->ymax << shift)) P_TOP[1] = (c->ymax << shift);
1179 if (P_TOPRIGHT[0] < (c->xmin << shift)) P_TOPRIGHT[0] = (c->xmin << shift);
1180 if (P_TOPRIGHT[1] > (c->ymax << shift)) P_TOPRIGHT[1] = (c->ymax << shift);
1182 P_MEDIAN[0] = mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
1183 P_MEDIAN[1] = mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
1185 c->pred_x = P_LEFT[0];
1186 c->pred_y = P_LEFT[1];
1188 if(mv_table == s->b_forw_mv_table){
1189 mv_scale= (s->pb_time<<16) / (s->pp_time<<shift);
1191 mv_scale= ((s->pb_time - s->pp_time)<<16) / (s->pp_time<<shift);
1194 dmin = ff_epzs_motion_search(s, &mx, &my, P, 0, ref_index, s->p_mv_table, mv_scale, 0, 16);
1199 dmin= c->sub_motion_search(s, &mx, &my, dmin, 0, ref_index, 0, 16);
1201 if(c->avctx->me_sub_cmp != c->avctx->mb_cmp && !c->skip)
1202 dmin= get_mb_score(s, mx, my, 0, ref_index, 0, 16, 1);
1204 // s->mb_type[mb_y*s->mb_width + mb_x]= mb_type;
1205 mv_table[mot_xy][0]= mx;
1206 mv_table[mot_xy][1]= my;
1211 static inline int check_bidir_mv(MpegEncContext * s,
1212 int motion_fx, int motion_fy,
1213 int motion_bx, int motion_by,
1214 int pred_fx, int pred_fy,
1215 int pred_bx, int pred_by,
1219 //FIXME better f_code prediction (max mv & distance)
1221 MotionEstContext * const c= &s->me;
1222 uint8_t * const mv_penalty_f= c->mv_penalty[s->f_code] + MAX_MV; // f_code of the prev frame
1223 uint8_t * const mv_penalty_b= c->mv_penalty[s->b_code] + MAX_MV; // f_code of the prev frame
1224 int stride= c->stride;
1225 uint8_t *dest_y = c->scratchpad;
1230 uint8_t **src_data= c->src[0];
1231 uint8_t **ref_data= c->ref[0];
1232 uint8_t **ref2_data= c->ref[2];
1234 if(s->quarter_sample){
1235 dxy = ((motion_fy & 3) << 2) | (motion_fx & 3);
1236 src_x = motion_fx >> 2;
1237 src_y = motion_fy >> 2;
1239 ptr = ref_data[0] + (src_y * stride) + src_x;
1240 s->qdsp.put_qpel_pixels_tab[0][dxy](dest_y, ptr, stride);
1242 dxy = ((motion_by & 3) << 2) | (motion_bx & 3);
1243 src_x = motion_bx >> 2;
1244 src_y = motion_by >> 2;
1246 ptr = ref2_data[0] + (src_y * stride) + src_x;
1247 s->qdsp.avg_qpel_pixels_tab[size][dxy](dest_y, ptr, stride);
1249 dxy = ((motion_fy & 1) << 1) | (motion_fx & 1);
1250 src_x = motion_fx >> 1;
1251 src_y = motion_fy >> 1;
1253 ptr = ref_data[0] + (src_y * stride) + src_x;
1254 s->hdsp.put_pixels_tab[size][dxy](dest_y , ptr , stride, h);
1256 dxy = ((motion_by & 1) << 1) | (motion_bx & 1);
1257 src_x = motion_bx >> 1;
1258 src_y = motion_by >> 1;
1260 ptr = ref2_data[0] + (src_y * stride) + src_x;
1261 s->hdsp.avg_pixels_tab[size][dxy](dest_y , ptr , stride, h);
1264 fbmin = (mv_penalty_f[motion_fx-pred_fx] + mv_penalty_f[motion_fy-pred_fy])*c->mb_penalty_factor
1265 +(mv_penalty_b[motion_bx-pred_bx] + mv_penalty_b[motion_by-pred_by])*c->mb_penalty_factor
1266 + s->mecc.mb_cmp[size](s, src_data[0], dest_y, stride, h); // FIXME new_pic
1268 if(c->avctx->mb_cmp&FF_CMP_CHROMA){
1275 /* refine the bidir vectors in hq mode and return the score in both lq & hq mode*/
1276 static inline int bidir_refine(MpegEncContext * s, int mb_x, int mb_y)
1278 MotionEstContext * const c= &s->me;
1279 const int mot_stride = s->mb_stride;
1280 const int xy = mb_y *mot_stride + mb_x;
1282 int pred_fx= s->b_bidir_forw_mv_table[xy-1][0];
1283 int pred_fy= s->b_bidir_forw_mv_table[xy-1][1];
1284 int pred_bx= s->b_bidir_back_mv_table[xy-1][0];
1285 int pred_by= s->b_bidir_back_mv_table[xy-1][1];
1286 int motion_fx= s->b_bidir_forw_mv_table[xy][0]= s->b_forw_mv_table[xy][0];
1287 int motion_fy= s->b_bidir_forw_mv_table[xy][1]= s->b_forw_mv_table[xy][1];
1288 int motion_bx= s->b_bidir_back_mv_table[xy][0]= s->b_back_mv_table[xy][0];
1289 int motion_by= s->b_bidir_back_mv_table[xy][1]= s->b_back_mv_table[xy][1];
1290 const int flags= c->sub_flags;
1291 const int qpel= flags&FLAG_QPEL;
1292 const int shift= 1+qpel;
1293 const int xmin= c->xmin<<shift;
1294 const int ymin= c->ymin<<shift;
1295 const int xmax= c->xmax<<shift;
1296 const int ymax= c->ymax<<shift;
1297 #define HASH(fx,fy,bx,by) ((fx)+17*(fy)+63*(bx)+117*(by))
1298 #define HASH8(fx,fy,bx,by) ((uint8_t)HASH(fx,fy,bx,by))
1299 int hashidx= HASH(motion_fx,motion_fy, motion_bx, motion_by);
1300 uint8_t map[256] = { 0 };
1302 map[hashidx&255] = 1;
1304 fbmin= check_bidir_mv(s, motion_fx, motion_fy,
1305 motion_bx, motion_by,
1310 if(s->avctx->bidir_refine){
1312 static const uint8_t limittab[5]={0,8,32,64,80};
1313 const int limit= limittab[s->avctx->bidir_refine];
1314 static const int8_t vect[][4]={
1315 { 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},
1317 { 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},
1318 { 0, 1, 0, 1}, { 0,-1, 0,-1}, { 1, 0, 1, 0}, {-1, 0,-1, 0},
1319 { 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},
1320 { 0,-1, 0, 1}, { 0, 1, 0,-1}, {-1, 0, 1, 0}, { 1, 0,-1, 0},
1322 { 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},
1323 { 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},
1324 { 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},
1325 { 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},
1327 { 1, 1, 1, 1}, {-1,-1,-1,-1},
1328 { 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},
1329 { 1, 1,-1,-1}, {-1,-1, 1, 1}, { 1,-1,-1, 1}, {-1, 1, 1,-1}, { 1,-1, 1,-1}, {-1, 1,-1, 1},
1331 static const uint8_t hash[]={
1332 HASH8( 0, 0, 0, 1), HASH8( 0, 0, 0,-1), HASH8( 0, 0, 1, 0), HASH8( 0, 0,-1, 0), HASH8( 0, 1, 0, 0), HASH8( 0,-1, 0, 0), HASH8( 1, 0, 0, 0), HASH8(-1, 0, 0, 0),
1334 HASH8( 0, 0, 1, 1), HASH8( 0, 0,-1,-1), HASH8( 0, 1, 1, 0), HASH8( 0,-1,-1, 0), HASH8( 1, 1, 0, 0), HASH8(-1,-1, 0, 0), HASH8( 1, 0, 0, 1), HASH8(-1, 0, 0,-1),
1335 HASH8( 0, 1, 0, 1), HASH8( 0,-1, 0,-1), HASH8( 1, 0, 1, 0), HASH8(-1, 0,-1, 0),
1336 HASH8( 0, 0,-1, 1), HASH8( 0, 0, 1,-1), HASH8( 0,-1, 1, 0), HASH8( 0, 1,-1, 0), HASH8(-1, 1, 0, 0), HASH8( 1,-1, 0, 0), HASH8( 1, 0, 0,-1), HASH8(-1, 0, 0, 1),
1337 HASH8( 0,-1, 0, 1), HASH8( 0, 1, 0,-1), HASH8(-1, 0, 1, 0), HASH8( 1, 0,-1, 0),
1339 HASH8( 0, 1, 1, 1), HASH8( 0,-1,-1,-1), HASH8( 1, 1, 1, 0), HASH8(-1,-1,-1, 0), HASH8( 1, 1, 0, 1), HASH8(-1,-1, 0,-1), HASH8( 1, 0, 1, 1), HASH8(-1, 0,-1,-1),
1340 HASH8( 0,-1, 1, 1), HASH8( 0, 1,-1,-1), HASH8(-1, 1, 1, 0), HASH8( 1,-1,-1, 0), HASH8( 1, 1, 0,-1), HASH8(-1,-1, 0, 1), HASH8( 1, 0,-1, 1), HASH8(-1, 0, 1,-1),
1341 HASH8( 0, 1,-1, 1), HASH8( 0,-1, 1,-1), HASH8( 1,-1, 1, 0), HASH8(-1, 1,-1, 0), HASH8(-1, 1, 0, 1), HASH8( 1,-1, 0,-1), HASH8( 1, 0, 1,-1), HASH8(-1, 0,-1, 1),
1342 HASH8( 0, 1, 1,-1), HASH8( 0,-1,-1, 1), HASH8( 1, 1,-1, 0), HASH8(-1,-1, 1, 0), HASH8( 1,-1, 0, 1), HASH8(-1, 1, 0,-1), HASH8(-1, 0, 1, 1), HASH8( 1, 0,-1,-1),
1344 HASH8( 1, 1, 1, 1), HASH8(-1,-1,-1,-1),
1345 HASH8( 1, 1, 1,-1), HASH8(-1,-1,-1, 1), HASH8( 1, 1,-1, 1), HASH8(-1,-1, 1,-1), HASH8( 1,-1, 1, 1), HASH8(-1, 1,-1,-1), HASH8(-1, 1, 1, 1), HASH8( 1,-1,-1,-1),
1346 HASH8( 1, 1,-1,-1), HASH8(-1,-1, 1, 1), HASH8( 1,-1,-1, 1), HASH8(-1, 1, 1,-1), HASH8( 1,-1, 1,-1), HASH8(-1, 1,-1, 1),
1349 #define CHECK_BIDIR(fx,fy,bx,by)\
1350 if( !map[(hashidx+HASH(fx,fy,bx,by))&255]\
1351 &&(fx<=0 || motion_fx+fx<=xmax) && (fy<=0 || motion_fy+fy<=ymax) && (bx<=0 || motion_bx+bx<=xmax) && (by<=0 || motion_by+by<=ymax)\
1352 &&(fx>=0 || motion_fx+fx>=xmin) && (fy>=0 || motion_fy+fy>=ymin) && (bx>=0 || motion_bx+bx>=xmin) && (by>=0 || motion_by+by>=ymin)){\
1354 map[(hashidx+HASH(fx,fy,bx,by))&255] = 1;\
1355 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);\
1357 hashidx += HASH(fx,fy,bx,by);\
1366 #define CHECK_BIDIR2(a,b,c,d)\
1367 CHECK_BIDIR(a,b,c,d)\
1368 CHECK_BIDIR(-(a),-(b),-(c),-(d))
1375 CHECK_BIDIR2(0,0,0,1)
1376 CHECK_BIDIR2(0,0,1,0)
1377 CHECK_BIDIR2(0,1,0,0)
1378 CHECK_BIDIR2(1,0,0,0)
1380 for(i=8; i<limit; i++){
1381 int fx= motion_fx+vect[i][0];
1382 int fy= motion_fy+vect[i][1];
1383 int bx= motion_bx+vect[i][2];
1384 int by= motion_by+vect[i][3];
1386 int a= (xmax - FFMAX(fx,bx))|(FFMIN(fx,bx) - xmin);
1387 int b= (ymax - FFMAX(fy,by))|(FFMIN(fy,by) - ymin);
1389 map[(hashidx+hash[i])&255] = 1;
1391 if(!map[(hashidx+hash[i])&255]){
1393 map[(hashidx+hash[i])&255] = 1;
1394 score= check_bidir_mv(s, fx, fy, bx, by, pred_fx, pred_fy, pred_bx, pred_by, 0, 16);
1405 int a= FFMIN(xmax - FFMAX(fx,bx), FFMIN(fx,bx) - xmin);
1406 int b= FFMIN(ymax - FFMAX(fy,by), FFMIN(fy,by) - ymin);
1407 borderdist= FFMIN(a,b);
1415 s->b_bidir_forw_mv_table[xy][0]= motion_fx;
1416 s->b_bidir_forw_mv_table[xy][1]= motion_fy;
1417 s->b_bidir_back_mv_table[xy][0]= motion_bx;
1418 s->b_bidir_back_mv_table[xy][1]= motion_by;
1423 static inline int direct_search(MpegEncContext * s, int mb_x, int mb_y)
1425 MotionEstContext * const c= &s->me;
1427 const int mot_stride = s->mb_stride;
1428 const int mot_xy = mb_y*mot_stride + mb_x;
1429 const int shift= 1+s->quarter_sample;
1431 const int time_pp= s->pp_time;
1432 const int time_pb= s->pb_time;
1433 int mx, my, xmin, xmax, ymin, ymax;
1434 int16_t (*mv_table)[2]= s->b_direct_mv_table;
1436 c->current_mv_penalty= c->mv_penalty[1] + MAX_MV;
1437 ymin= xmin=(-32)>>shift;
1438 ymax= xmax= 31>>shift;
1440 if (IS_8X8(s->next_picture.mb_type[mot_xy])) {
1441 s->mv_type= MV_TYPE_8X8;
1443 s->mv_type= MV_TYPE_16X16;
1447 int index= s->block_index[i];
1450 c->co_located_mv[i][0] = s->next_picture.motion_val[0][index][0];
1451 c->co_located_mv[i][1] = s->next_picture.motion_val[0][index][1];
1452 c->direct_basis_mv[i][0]= c->co_located_mv[i][0]*time_pb/time_pp + ((i& 1)<<(shift+3));
1453 c->direct_basis_mv[i][1]= c->co_located_mv[i][1]*time_pb/time_pp + ((i>>1)<<(shift+3));
1454 // c->direct_basis_mv[1][i][0]= c->co_located_mv[i][0]*(time_pb - time_pp)/time_pp + ((i &1)<<(shift+3);
1455 // c->direct_basis_mv[1][i][1]= c->co_located_mv[i][1]*(time_pb - time_pp)/time_pp + ((i>>1)<<(shift+3);
1457 max= FFMAX(c->direct_basis_mv[i][0], c->direct_basis_mv[i][0] - c->co_located_mv[i][0])>>shift;
1458 min= FFMIN(c->direct_basis_mv[i][0], c->direct_basis_mv[i][0] - c->co_located_mv[i][0])>>shift;
1459 max+= 16*mb_x + 1; // +-1 is for the simpler rounding
1461 xmax= FFMIN(xmax, s->width - max);
1462 xmin= FFMAX(xmin, - 16 - min);
1464 max= FFMAX(c->direct_basis_mv[i][1], c->direct_basis_mv[i][1] - c->co_located_mv[i][1])>>shift;
1465 min= FFMIN(c->direct_basis_mv[i][1], c->direct_basis_mv[i][1] - c->co_located_mv[i][1])>>shift;
1466 max+= 16*mb_y + 1; // +-1 is for the simpler rounding
1468 ymax= FFMIN(ymax, s->height - max);
1469 ymin= FFMAX(ymin, - 16 - min);
1471 if(s->mv_type == MV_TYPE_16X16) break;
1474 av_assert2(xmax <= 15 && ymax <= 15 && xmin >= -16 && ymin >= -16);
1476 if(xmax < 0 || xmin >0 || ymax < 0 || ymin > 0){
1477 s->b_direct_mv_table[mot_xy][0]= 0;
1478 s->b_direct_mv_table[mot_xy][1]= 0;
1480 return 256*256*256*64;
1487 c->flags |= FLAG_DIRECT;
1488 c->sub_flags |= FLAG_DIRECT;
1492 P_LEFT[0] = av_clip(mv_table[mot_xy - 1][0], xmin<<shift, xmax<<shift);
1493 P_LEFT[1] = av_clip(mv_table[mot_xy - 1][1], ymin<<shift, ymax<<shift);
1495 /* special case for first line */
1496 if (!s->first_slice_line) { //FIXME maybe allow this over thread boundary as it is clipped
1497 P_TOP[0] = av_clip(mv_table[mot_xy - mot_stride ][0], xmin<<shift, xmax<<shift);
1498 P_TOP[1] = av_clip(mv_table[mot_xy - mot_stride ][1], ymin<<shift, ymax<<shift);
1499 P_TOPRIGHT[0] = av_clip(mv_table[mot_xy - mot_stride + 1 ][0], xmin<<shift, xmax<<shift);
1500 P_TOPRIGHT[1] = av_clip(mv_table[mot_xy - mot_stride + 1 ][1], ymin<<shift, ymax<<shift);
1502 P_MEDIAN[0]= mid_pred(P_LEFT[0], P_TOP[0], P_TOPRIGHT[0]);
1503 P_MEDIAN[1]= mid_pred(P_LEFT[1], P_TOP[1], P_TOPRIGHT[1]);
1506 dmin = ff_epzs_motion_search(s, &mx, &my, P, 0, 0, mv_table, 1<<(16-shift), 0, 16);
1507 if(c->sub_flags&FLAG_QPEL)
1508 dmin = qpel_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
1510 dmin = hpel_motion_search(s, &mx, &my, dmin, 0, 0, 0, 16);
1512 if(c->avctx->me_sub_cmp != c->avctx->mb_cmp && !c->skip)
1513 dmin= get_mb_score(s, mx, my, 0, 0, 0, 16, 1);
1515 get_limits(s, 16*mb_x, 16*mb_y); //restore c->?min/max, maybe not needed
1517 mv_table[mot_xy][0]= mx;
1518 mv_table[mot_xy][1]= my;
1519 c->flags &= ~FLAG_DIRECT;
1520 c->sub_flags &= ~FLAG_DIRECT;
1525 void ff_estimate_b_frame_motion(MpegEncContext * s,
1528 MotionEstContext * const c= &s->me;
1529 const int penalty_factor= c->mb_penalty_factor;
1530 int fmin, bmin, dmin, fbmin, bimin, fimin;
1532 const int xy = mb_y*s->mb_stride + mb_x;
1533 init_ref(c, s->new_picture.f->data, s->last_picture.f->data,
1534 s->next_picture.f->data, 16 * mb_x, 16 * mb_y, 2);
1536 get_limits(s, 16*mb_x, 16*mb_y);
1540 if (s->codec_id == AV_CODEC_ID_MPEG4 && s->next_picture.mbskip_table[xy]) {
1541 int score= direct_search(s, mb_x, mb_y); //FIXME just check 0,0
1543 score= ((unsigned)(score*score + 128*256))>>16;
1544 c->mc_mb_var_sum_temp += score;
1545 s->current_picture.mc_mb_var[mb_y*s->mb_stride + mb_x] = score; //FIXME use SSE
1546 s->mb_type[mb_y*s->mb_stride + mb_x]= CANDIDATE_MB_TYPE_DIRECT0;
1551 if (s->codec_id == AV_CODEC_ID_MPEG4)
1552 dmin= direct_search(s, mb_x, mb_y);
1555 //FIXME penalty stuff for non mpeg4
1557 fmin = estimate_motion_b(s, mb_x, mb_y, s->b_forw_mv_table, 0, s->f_code) +
1561 bmin = estimate_motion_b(s, mb_x, mb_y, s->b_back_mv_table, 2, s->b_code) +
1563 av_dlog(s, " %d %d ", s->b_forw_mv_table[xy][0], s->b_forw_mv_table[xy][1]);
1566 fbmin= bidir_refine(s, mb_x, mb_y) + penalty_factor;
1567 av_dlog(s, "%d %d %d %d\n", dmin, fmin, bmin, fbmin);
1569 if(s->flags & CODEC_FLAG_INTERLACED_ME){
1570 //FIXME mb type penalty
1572 c->current_mv_penalty= c->mv_penalty[s->f_code] + MAX_MV;
1573 fimin= interlaced_search(s, 0,
1574 s->b_field_mv_table[0], s->b_field_select_table[0],
1575 s->b_forw_mv_table[xy][0], s->b_forw_mv_table[xy][1], 0);
1576 c->current_mv_penalty= c->mv_penalty[s->b_code] + MAX_MV;
1577 bimin= interlaced_search(s, 2,
1578 s->b_field_mv_table[1], s->b_field_select_table[1],
1579 s->b_back_mv_table[xy][0], s->b_back_mv_table[xy][1], 0);
1581 fimin= bimin= INT_MAX;
1585 type = CANDIDATE_MB_TYPE_FORWARD;
1589 type = CANDIDATE_MB_TYPE_DIRECT;
1593 type= CANDIDATE_MB_TYPE_BACKWARD;
1597 type= CANDIDATE_MB_TYPE_BIDIR;
1601 type= CANDIDATE_MB_TYPE_FORWARD_I;
1605 type= CANDIDATE_MB_TYPE_BACKWARD_I;
1608 score= ((unsigned)(score*score + 128*256))>>16;
1609 c->mc_mb_var_sum_temp += score;
1610 s->current_picture.mc_mb_var[mb_y*s->mb_stride + mb_x] = score; //FIXME use SSE
1613 if(c->avctx->mb_decision > FF_MB_DECISION_SIMPLE){
1614 type= CANDIDATE_MB_TYPE_FORWARD | CANDIDATE_MB_TYPE_BACKWARD | CANDIDATE_MB_TYPE_BIDIR | CANDIDATE_MB_TYPE_DIRECT;
1616 type |= CANDIDATE_MB_TYPE_FORWARD_I;
1618 type |= CANDIDATE_MB_TYPE_BACKWARD_I;
1619 if(fimin < INT_MAX && bimin < INT_MAX){
1620 type |= CANDIDATE_MB_TYPE_BIDIR_I;
1622 //FIXME something smarter
1623 if(dmin>256*256*16) type&= ~CANDIDATE_MB_TYPE_DIRECT; //do not try direct mode if it is invalid for this MB
1624 if (s->codec_id == AV_CODEC_ID_MPEG4 && type&CANDIDATE_MB_TYPE_DIRECT &&
1625 s->mpv_flags & FF_MPV_FLAG_MV0 && *(uint32_t*)s->b_direct_mv_table[xy])
1626 type |= CANDIDATE_MB_TYPE_DIRECT0;
1629 s->mb_type[mb_y*s->mb_stride + mb_x]= type;
1632 /* find best f_code for ME which do unlimited searches */
1633 int ff_get_best_fcode(MpegEncContext * s, int16_t (*mv_table)[2], int type)
1635 if(s->me_method>=ME_EPZS){
1637 int i, y, range= s->avctx->me_range ? s->avctx->me_range : (INT_MAX/2);
1638 uint8_t * fcode_tab= s->fcode_tab;
1640 int best_score=-10000000;
1642 if(s->msmpeg4_version)
1643 range= FFMIN(range, 16);
1644 else if(s->codec_id == AV_CODEC_ID_MPEG2VIDEO && s->avctx->strict_std_compliance >= FF_COMPLIANCE_NORMAL)
1645 range= FFMIN(range, 256);
1647 for(i=0; i<8; i++) score[i]= s->mb_num*(8-i);
1649 for(y=0; y<s->mb_height; y++){
1651 int xy= y*s->mb_stride;
1652 for(x=0; x<s->mb_width; x++){
1653 if(s->mb_type[xy] & type){
1654 int mx= mv_table[xy][0];
1655 int my= mv_table[xy][1];
1656 int fcode= FFMAX(fcode_tab[mx + MAX_MV],
1657 fcode_tab[my + MAX_MV]);
1660 if(mx >= range || mx < -range ||
1661 my >= range || my < -range)
1664 for(j=0; j<fcode && j<8; j++){
1665 if(s->pict_type==AV_PICTURE_TYPE_B || s->current_picture.mc_mb_var[xy] < s->current_picture.mb_var[xy])
1674 if(score[i] > best_score){
1675 best_score= score[i];
1686 void ff_fix_long_p_mvs(MpegEncContext * s)
1688 MotionEstContext * const c= &s->me;
1689 const int f_code= s->f_code;
1691 av_assert0(s->pict_type==AV_PICTURE_TYPE_P);
1693 range = (((s->out_format == FMT_MPEG1 || s->msmpeg4_version) ? 8 : 16) << f_code);
1695 av_assert0(range <= 16 || !s->msmpeg4_version);
1696 av_assert0(range <=256 || !(s->codec_id == AV_CODEC_ID_MPEG2VIDEO && s->avctx->strict_std_compliance >= FF_COMPLIANCE_NORMAL));
1698 if(c->avctx->me_range && range > c->avctx->me_range) range= c->avctx->me_range;
1700 if(s->flags&CODEC_FLAG_4MV){
1701 const int wrap= s->b8_stride;
1703 /* clip / convert to intra 8x8 type MVs */
1704 for(y=0; y<s->mb_height; y++){
1706 int i= y*s->mb_stride;
1709 for(x=0; x<s->mb_width; x++){
1710 if(s->mb_type[i]&CANDIDATE_MB_TYPE_INTER4V){
1712 for(block=0; block<4; block++){
1713 int off= (block& 1) + (block>>1)*wrap;
1714 int mx = s->current_picture.motion_val[0][ xy + off ][0];
1715 int my = s->current_picture.motion_val[0][ xy + off ][1];
1717 if( mx >=range || mx <-range
1718 || my >=range || my <-range){
1719 s->mb_type[i] &= ~CANDIDATE_MB_TYPE_INTER4V;
1720 s->mb_type[i] |= CANDIDATE_MB_TYPE_INTRA;
1721 s->current_picture.mb_type[i] = CANDIDATE_MB_TYPE_INTRA;
1734 * @param truncate 1 for truncation, 0 for using intra
1736 void ff_fix_long_mvs(MpegEncContext * s, uint8_t *field_select_table, int field_select,
1737 int16_t (*mv_table)[2], int f_code, int type, int truncate)
1739 MotionEstContext * const c= &s->me;
1740 int y, h_range, v_range;
1742 // RAL: 8 in MPEG-1, 16 in MPEG-4
1743 int range = (((s->out_format == FMT_MPEG1 || s->msmpeg4_version) ? 8 : 16) << f_code);
1745 if(c->avctx->me_range && range > c->avctx->me_range) range= c->avctx->me_range;
1748 v_range= field_select_table ? range>>1 : range;
1750 /* clip / convert to intra 16x16 type MVs */
1751 for(y=0; y<s->mb_height; y++){
1753 int xy= y*s->mb_stride;
1754 for(x=0; x<s->mb_width; x++){
1755 if (s->mb_type[xy] & type){ // RAL: "type" test added...
1756 if (!field_select_table || field_select_table[xy] == field_select) {
1757 if( mv_table[xy][0] >=h_range || mv_table[xy][0] <-h_range
1758 || mv_table[xy][1] >=v_range || mv_table[xy][1] <-v_range){
1761 if (mv_table[xy][0] > h_range-1) mv_table[xy][0]= h_range-1;
1762 else if(mv_table[xy][0] < -h_range ) mv_table[xy][0]= -h_range;
1763 if (mv_table[xy][1] > v_range-1) mv_table[xy][1]= v_range-1;
1764 else if(mv_table[xy][1] < -v_range ) mv_table[xy][1]= -v_range;
1766 s->mb_type[xy] &= ~type;
1767 s->mb_type[xy] |= CANDIDATE_MB_TYPE_INTRA;