Commit | Line | Data |
---|---|---|
26b4fe82 AJ |
1 | /* |
2 | * H.26L/H.264/AVC/JVT/14496-10/... parser | |
3 | * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at> | |
4 | * | |
2912e87a | 5 | * This file is part of Libav. |
26b4fe82 | 6 | * |
2912e87a | 7 | * Libav is free software; you can redistribute it and/or |
26b4fe82 AJ |
8 | * modify it under the terms of the GNU Lesser General Public |
9 | * License as published by the Free Software Foundation; either | |
10 | * version 2.1 of the License, or (at your option) any later version. | |
11 | * | |
2912e87a | 12 | * Libav is distributed in the hope that it will be useful, |
26b4fe82 AJ |
13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
15 | * Lesser General Public License for more details. | |
16 | * | |
17 | * You should have received a copy of the GNU Lesser General Public | |
2912e87a | 18 | * License along with Libav; if not, write to the Free Software |
26b4fe82 | 19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
26b4fe82 AJ |
20 | */ |
21 | ||
22 | /** | |
ba87f080 | 23 | * @file |
26b4fe82 AJ |
24 | * H.264 / AVC / MPEG4 part10 parser. |
25 | * @author Michael Niedermayer <michaelni@gmx.at> | |
26 | */ | |
27 | ||
28 | #include "parser.h" | |
ff6474dd IS |
29 | #include "h264data.h" |
30 | #include "golomb.h" | |
26b4fe82 AJ |
31 | |
32 | #include <assert.h> | |
33 | ||
34 | ||
13eb6b90 | 35 | static int ff_h264_find_frame_end(H264Context *h, const uint8_t *buf, int buf_size) |
26b4fe82 AJ |
36 | { |
37 | int i; | |
38 | uint32_t state; | |
39 | ParseContext *pc = &(h->s.parse_context); | |
40 | //printf("first %02X%02X%02X%02X\n", buf[0], buf[1],buf[2],buf[3]); | |
41 | // mb_addr= pc->mb_addr - 1; | |
42 | state= pc->state; | |
43 | if(state>13) | |
44 | state= 7; | |
45 | ||
46 | for(i=0; i<buf_size; i++){ | |
47 | if(state==7){ | |
b250f9c6 | 48 | #if HAVE_FAST_UNALIGNED |
e4f1ec3a MN |
49 | /* we check i<buf_size instead of i+3/7 because its simpler |
50 | * and there should be FF_INPUT_BUFFER_PADDING_SIZE bytes at the end | |
51 | */ | |
b250f9c6 | 52 | # if HAVE_FAST_64BIT |
4d8eb2e8 | 53 | while(i<buf_size && !((~*(const uint64_t*)(buf+i) & (*(const uint64_t*)(buf+i) - 0x0101010101010101ULL)) & 0x8080808080808080ULL)) |
5cb5023c MN |
54 | i+=8; |
55 | # else | |
4d8eb2e8 | 56 | while(i<buf_size && !((~*(const uint32_t*)(buf+i) & (*(const uint32_t*)(buf+i) - 0x01010101U)) & 0x80808080U)) |
5cb5023c MN |
57 | i+=4; |
58 | # endif | |
59 | #endif | |
26b4fe82 AJ |
60 | for(; i<buf_size; i++){ |
61 | if(!buf[i]){ | |
62 | state=2; | |
63 | break; | |
64 | } | |
65 | } | |
66 | }else if(state<=2){ | |
67 | if(buf[i]==1) state^= 5; //2->7, 1->4, 0->5 | |
68 | else if(buf[i]) state = 7; | |
69 | else state>>=1; //2->1, 1->0, 0->0 | |
70 | }else if(state<=5){ | |
71 | int v= buf[i] & 0x1F; | |
9e85f9c5 | 72 | if(v==6 || v==7 || v==8 || v==9){ |
26b4fe82 AJ |
73 | if(pc->frame_start_found){ |
74 | i++; | |
9aa1cfec | 75 | goto found; |
26b4fe82 AJ |
76 | } |
77 | }else if(v==1 || v==2 || v==5){ | |
78 | if(pc->frame_start_found){ | |
79 | state+=8; | |
80 | continue; | |
81 | }else | |
82 | pc->frame_start_found = 1; | |
83 | } | |
84 | state= 7; | |
85 | }else{ | |
86 | if(buf[i] & 0x80) | |
87 | goto found; | |
88 | state= 7; | |
89 | } | |
90 | } | |
91 | pc->state= state; | |
92 | return END_NOT_FOUND; | |
9aa1cfec DP |
93 | |
94 | found: | |
95 | pc->state=7; | |
96 | pc->frame_start_found= 0; | |
97 | return i-(state&5); | |
26b4fe82 AJ |
98 | } |
99 | ||
ff6474dd IS |
100 | /*! |
101 | * Parse NAL units of found picture and decode some basic information. | |
102 | * | |
103 | * @param s parser context. | |
104 | * @param avctx codec context. | |
105 | * @param buf buffer with field/frame data. | |
106 | * @param buf_size size of the buffer. | |
107 | */ | |
108 | static inline int parse_nal_units(AVCodecParserContext *s, | |
109 | AVCodecContext *avctx, | |
110 | const uint8_t *buf, int buf_size) | |
111 | { | |
112 | H264Context *h = s->priv_data; | |
113 | const uint8_t *buf_end = buf + buf_size; | |
96c3da93 | 114 | unsigned int pps_id; |
ff6474dd | 115 | unsigned int slice_type; |
8fa0ae06 | 116 | int state = -1; |
ff6474dd IS |
117 | const uint8_t *ptr; |
118 | ||
119 | /* set some sane default values */ | |
975a1447 | 120 | s->pict_type = AV_PICTURE_TYPE_I; |
0ed260c7 | 121 | s->key_frame = 0; |
ff6474dd IS |
122 | |
123 | h->s.avctx= avctx; | |
0ed260c7 | 124 | h->sei_recovery_frame_cnt = -1; |
c733922e IS |
125 | h->sei_dpb_output_delay = 0; |
126 | h->sei_cpb_removal_delay = -1; | |
127 | h->sei_buffering_period_present = 0; | |
ff6474dd | 128 | |
9479415e BC |
129 | if (!buf_size) |
130 | return 0; | |
131 | ||
ff6474dd IS |
132 | for(;;) { |
133 | int src_length, dst_length, consumed; | |
134 | buf = ff_find_start_code(buf, buf_end, &state); | |
135 | if(buf >= buf_end) | |
136 | break; | |
137 | --buf; | |
138 | src_length = buf_end - buf; | |
139 | switch (state & 0x1f) { | |
140 | case NAL_SLICE: | |
141 | case NAL_IDR_SLICE: | |
142 | // Do not walk the whole buffer just to decode slice header | |
143 | if (src_length > 20) | |
144 | src_length = 20; | |
145 | break; | |
146 | } | |
147 | ptr= ff_h264_decode_nal(h, buf, &dst_length, &consumed, src_length); | |
148 | if (ptr==NULL || dst_length < 0) | |
149 | break; | |
150 | ||
151 | init_get_bits(&h->s.gb, ptr, 8*dst_length); | |
152 | switch(h->nal_unit_type) { | |
153 | case NAL_SPS: | |
154 | ff_h264_decode_seq_parameter_set(h); | |
155 | break; | |
156 | case NAL_PPS: | |
157 | ff_h264_decode_picture_parameter_set(h, h->s.gb.size_in_bits); | |
158 | break; | |
159 | case NAL_SEI: | |
160 | ff_h264_decode_sei(h); | |
161 | break; | |
162 | case NAL_IDR_SLICE: | |
0ed260c7 IS |
163 | s->key_frame = 1; |
164 | /* fall through */ | |
ff6474dd IS |
165 | case NAL_SLICE: |
166 | get_ue_golomb(&h->s.gb); // skip first_mb_in_slice | |
167 | slice_type = get_ue_golomb_31(&h->s.gb); | |
168 | s->pict_type = golomb_to_pict_type[slice_type % 5]; | |
0ed260c7 IS |
169 | if (h->sei_recovery_frame_cnt >= 0) { |
170 | /* key frame, since recovery_frame_cnt is set */ | |
171 | s->key_frame = 1; | |
172 | } | |
96c3da93 IS |
173 | pps_id= get_ue_golomb(&h->s.gb); |
174 | if(pps_id>=MAX_PPS_COUNT) { | |
175 | av_log(h->s.avctx, AV_LOG_ERROR, "pps_id out of range\n"); | |
176 | return -1; | |
177 | } | |
178 | if(!h->pps_buffers[pps_id]) { | |
179 | av_log(h->s.avctx, AV_LOG_ERROR, "non-existing PPS referenced\n"); | |
180 | return -1; | |
181 | } | |
182 | h->pps= *h->pps_buffers[pps_id]; | |
183 | if(!h->sps_buffers[h->pps.sps_id]) { | |
184 | av_log(h->s.avctx, AV_LOG_ERROR, "non-existing SPS referenced\n"); | |
185 | return -1; | |
186 | } | |
187 | h->sps = *h->sps_buffers[h->pps.sps_id]; | |
188 | h->frame_num = get_bits(&h->s.gb, h->sps.log2_max_frame_num); | |
189 | ||
fe9a3fbe | 190 | avctx->profile = ff_h264_get_profile(&h->sps); |
dd0cd3d2 RC |
191 | avctx->level = h->sps.level_idc; |
192 | ||
96c3da93 IS |
193 | if(h->sps.frame_mbs_only_flag){ |
194 | h->s.picture_structure= PICT_FRAME; | |
195 | }else{ | |
196 | if(get_bits1(&h->s.gb)) { //field_pic_flag | |
197 | h->s.picture_structure= PICT_TOP_FIELD + get_bits1(&h->s.gb); //bottom_field_flag | |
198 | } else { | |
199 | h->s.picture_structure= PICT_FRAME; | |
200 | } | |
201 | } | |
202 | ||
346db3ef IS |
203 | if(h->sps.pic_struct_present_flag) { |
204 | switch (h->sei_pic_struct) { | |
205 | case SEI_PIC_STRUCT_TOP_FIELD: | |
206 | case SEI_PIC_STRUCT_BOTTOM_FIELD: | |
fc9fe428 | 207 | s->repeat_pict = 0; |
346db3ef IS |
208 | break; |
209 | case SEI_PIC_STRUCT_FRAME: | |
210 | case SEI_PIC_STRUCT_TOP_BOTTOM: | |
211 | case SEI_PIC_STRUCT_BOTTOM_TOP: | |
fc9fe428 | 212 | s->repeat_pict = 1; |
346db3ef IS |
213 | break; |
214 | case SEI_PIC_STRUCT_TOP_BOTTOM_TOP: | |
215 | case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM: | |
fc9fe428 | 216 | s->repeat_pict = 2; |
346db3ef IS |
217 | break; |
218 | case SEI_PIC_STRUCT_FRAME_DOUBLING: | |
fc9fe428 | 219 | s->repeat_pict = 3; |
346db3ef IS |
220 | break; |
221 | case SEI_PIC_STRUCT_FRAME_TRIPLING: | |
fc9fe428 | 222 | s->repeat_pict = 5; |
346db3ef IS |
223 | break; |
224 | default: | |
fc9fe428 | 225 | s->repeat_pict = h->s.picture_structure == PICT_FRAME ? 1 : 0; |
346db3ef IS |
226 | break; |
227 | } | |
228 | } else { | |
fc9fe428 | 229 | s->repeat_pict = h->s.picture_structure == PICT_FRAME ? 1 : 0; |
346db3ef IS |
230 | } |
231 | ||
ff6474dd IS |
232 | return 0; /* no need to evaluate the rest */ |
233 | } | |
234 | buf += consumed; | |
235 | } | |
236 | /* didn't find a picture! */ | |
237 | av_log(h->s.avctx, AV_LOG_ERROR, "missing picture in access unit\n"); | |
238 | return -1; | |
239 | } | |
240 | ||
26b4fe82 AJ |
241 | static int h264_parse(AVCodecParserContext *s, |
242 | AVCodecContext *avctx, | |
243 | const uint8_t **poutbuf, int *poutbuf_size, | |
244 | const uint8_t *buf, int buf_size) | |
245 | { | |
246 | H264Context *h = s->priv_data; | |
247 | ParseContext *pc = &h->s.parse_context; | |
248 | int next; | |
249 | ||
82f1ffc7 HC |
250 | if (!h->got_first) { |
251 | h->got_first = 1; | |
23584bec HC |
252 | if (avctx->extradata_size) { |
253 | h->s.avctx = avctx; | |
254 | ff_h264_decode_extradata(h); | |
255 | } | |
256 | } | |
257 | ||
26b4fe82 AJ |
258 | if(s->flags & PARSER_FLAG_COMPLETE_FRAMES){ |
259 | next= buf_size; | |
260 | }else{ | |
261 | next= ff_h264_find_frame_end(h, buf, buf_size); | |
262 | ||
263 | if (ff_combine_frame(pc, next, &buf, &buf_size) < 0) { | |
264 | *poutbuf = NULL; | |
265 | *poutbuf_size = 0; | |
266 | return buf_size; | |
267 | } | |
268 | ||
269 | if(next<0 && next != END_NOT_FOUND){ | |
270 | assert(pc->last_index + next >= 0 ); | |
271 | ff_h264_find_frame_end(h, &pc->buffer[pc->last_index + next], -next); //update state | |
272 | } | |
a26ce1e2 | 273 | } |
ff6474dd IS |
274 | |
275 | parse_nal_units(s, avctx, buf, buf_size); | |
c733922e | 276 | |
2c0c5e12 | 277 | if (h->sei_cpb_removal_delay >= 0) { |
26aedb4a CEH |
278 | s->dts_sync_point = h->sei_buffering_period_present; |
279 | s->dts_ref_dts_delta = h->sei_cpb_removal_delay; | |
280 | s->pts_dts_delta = h->sei_dpb_output_delay; | |
2c0c5e12 IS |
281 | } else { |
282 | s->dts_sync_point = INT_MIN; | |
283 | s->dts_ref_dts_delta = INT_MIN; | |
284 | s->pts_dts_delta = INT_MIN; | |
285 | } | |
74a6df59 AC |
286 | if (s->flags & PARSER_FLAG_ONCE) { |
287 | s->flags &= PARSER_FLAG_COMPLETE_FRAMES; | |
288 | } | |
26b4fe82 AJ |
289 | |
290 | *poutbuf = buf; | |
291 | *poutbuf_size = buf_size; | |
292 | return next; | |
293 | } | |
294 | ||
295 | static int h264_split(AVCodecContext *avctx, | |
296 | const uint8_t *buf, int buf_size) | |
297 | { | |
298 | int i; | |
299 | uint32_t state = -1; | |
300 | int has_sps= 0; | |
301 | ||
302 | for(i=0; i<=buf_size; i++){ | |
303 | if((state&0xFFFFFF1F) == 0x107) | |
304 | has_sps=1; | |
305 | /* if((state&0xFFFFFF1F) == 0x101 || (state&0xFFFFFF1F) == 0x102 || (state&0xFFFFFF1F) == 0x105){ | |
306 | }*/ | |
307 | if((state&0xFFFFFF00) == 0x100 && (state&0xFFFFFF1F) != 0x107 && (state&0xFFFFFF1F) != 0x108 && (state&0xFFFFFF1F) != 0x109){ | |
308 | if(has_sps){ | |
309 | while(i>4 && buf[i-5]==0) i--; | |
310 | return i-4; | |
311 | } | |
312 | } | |
313 | if (i<buf_size) | |
314 | state= (state<<8) | buf[i]; | |
315 | } | |
316 | return 0; | |
317 | } | |
318 | ||
dd990075 | 319 | static void close(AVCodecParserContext *s) |
3ee4f5e4 MN |
320 | { |
321 | H264Context *h = s->priv_data; | |
322 | ParseContext *pc = &h->s.parse_context; | |
323 | ||
324 | av_free(pc->buffer); | |
15861962 | 325 | ff_h264_free_context(h); |
3ee4f5e4 MN |
326 | } |
327 | ||
e9ca315d RC |
328 | static int init(AVCodecParserContext *s) |
329 | { | |
330 | H264Context *h = s->priv_data; | |
331 | h->thread_context[0] = h; | |
332 | return 0; | |
333 | } | |
26b4fe82 | 334 | |
d36beb3f | 335 | AVCodecParser ff_h264_parser = { |
0d3d172f | 336 | { CODEC_ID_H264 }, |
26b4fe82 | 337 | sizeof(H264Context), |
e9ca315d | 338 | init, |
26b4fe82 | 339 | h264_parse, |
3ee4f5e4 | 340 | close, |
26b4fe82 AJ |
341 | h264_split, |
342 | }; |