]> git.sesse.net Git - ffmpeg/blob - libavcodec/libschroedingerenc.c
cavsdec: check dimensions being valid.
[ffmpeg] / libavcodec / libschroedingerenc.c
1 /*
2  * Dirac encoder support via Schroedinger libraries
3  * Copyright (c) 2008 BBC, Anuradha Suraparaju <asuraparaju at gmail dot com >
4  *
5  * This file is part of FFmpeg.
6  *
7  * FFmpeg is free software; you can redistribute it and/or
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  *
12  * FFmpeg is distributed in the hope that it will be useful,
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
18  * License along with FFmpeg; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21
22 /**
23 * @file
24 * Dirac encoder support via libschroedinger-1.0 libraries. More details about
25 * the Schroedinger project can be found at http://www.diracvideo.org/.
26 * The library implements Dirac Specification Version 2.2
27 * (http://dirac.sourceforge.net/specification.html).
28 */
29
30 #include <schroedinger/schro.h>
31 #include <schroedinger/schrodebug.h>
32 #include <schroedinger/schrovideoformat.h>
33
34 #include "libavutil/avassert.h"
35 #include "avcodec.h"
36 #include "internal.h"
37 #include "libschroedinger.h"
38 #include "bytestream.h"
39
40
41 /** libschroedinger encoder private data */
42 typedef struct SchroEncoderParams {
43     /** Schroedinger video format */
44     SchroVideoFormat *format;
45
46     /** Schroedinger frame format */
47     SchroFrameFormat frame_format;
48
49     /** frame being encoded */
50     AVFrame picture;
51
52     /** frame size */
53     int frame_size;
54
55     /** Schroedinger encoder handle*/
56     SchroEncoder* encoder;
57
58     /** buffer to store encoder output before writing it to the frame queue*/
59     unsigned char *enc_buf;
60
61     /** Size of encoder buffer*/
62     int enc_buf_size;
63
64     /** queue storing encoded frames */
65     FFSchroQueue enc_frame_queue;
66
67     /** end of sequence signalled */
68     int eos_signalled;
69
70     /** end of sequence pulled */
71     int eos_pulled;
72
73     /* counter for frames submitted to encoder, used as dts */
74     int64_t dts;
75 } SchroEncoderParams;
76
77 /**
78 * Works out Schro-compatible chroma format.
79 */
80 static int set_chroma_format(AVCodecContext *avccontext)
81 {
82     int num_formats = sizeof(schro_pixel_format_map) /
83                       sizeof(schro_pixel_format_map[0]);
84     int idx;
85
86     SchroEncoderParams *p_schro_params = avccontext->priv_data;
87
88     for (idx = 0; idx < num_formats; ++idx) {
89         if (schro_pixel_format_map[idx].ff_pix_fmt ==
90             avccontext->pix_fmt) {
91             p_schro_params->format->chroma_format =
92                             schro_pixel_format_map[idx].schro_pix_fmt;
93             return 0;
94         }
95     }
96
97     av_log(avccontext, AV_LOG_ERROR,
98            "This codec currently only supports planar YUV 4:2:0, 4:2:2"
99            " and 4:4:4 formats.\n");
100
101     return -1;
102 }
103
104 static int libschroedinger_encode_init(AVCodecContext *avccontext)
105 {
106     SchroEncoderParams *p_schro_params = avccontext->priv_data;
107     SchroVideoFormatEnum preset;
108
109     /* Initialize the libraries that libschroedinger depends on. */
110     schro_init();
111
112     /* Create an encoder object. */
113     p_schro_params->encoder = schro_encoder_new();
114
115     if (!p_schro_params->encoder) {
116         av_log(avccontext, AV_LOG_ERROR,
117                "Unrecoverable Error: schro_encoder_new failed. ");
118         return -1;
119     }
120
121     /* Initialize the format. */
122     preset = ff_get_schro_video_format_preset(avccontext);
123     p_schro_params->format =
124                     schro_encoder_get_video_format(p_schro_params->encoder);
125     schro_video_format_set_std_video_format(p_schro_params->format, preset);
126     p_schro_params->format->width  = avccontext->width;
127     p_schro_params->format->height = avccontext->height;
128
129     if (set_chroma_format(avccontext) == -1)
130         return -1;
131
132     if (avccontext->color_primaries == AVCOL_PRI_BT709) {
133         p_schro_params->format->colour_primaries = SCHRO_COLOUR_PRIMARY_HDTV;
134     } else if (avccontext->color_primaries == AVCOL_PRI_BT470BG) {
135         p_schro_params->format->colour_primaries = SCHRO_COLOUR_PRIMARY_SDTV_625;
136     } else if (avccontext->color_primaries == AVCOL_PRI_SMPTE170M) {
137         p_schro_params->format->colour_primaries = SCHRO_COLOUR_PRIMARY_SDTV_525;
138     }
139
140     if (avccontext->colorspace == AVCOL_SPC_BT709) {
141         p_schro_params->format->colour_matrix = SCHRO_COLOUR_MATRIX_HDTV;
142     } else if (avccontext->colorspace == AVCOL_SPC_BT470BG) {
143         p_schro_params->format->colour_matrix = SCHRO_COLOUR_MATRIX_SDTV;
144     }
145
146     if (avccontext->color_trc == AVCOL_TRC_BT709) {
147         p_schro_params->format->transfer_function = SCHRO_TRANSFER_CHAR_TV_GAMMA;
148     }
149
150     if (ff_get_schro_frame_format(p_schro_params->format->chroma_format,
151                                   &p_schro_params->frame_format) == -1) {
152         av_log(avccontext, AV_LOG_ERROR,
153                "This codec currently supports only planar YUV 4:2:0, 4:2:2"
154                " and 4:4:4 formats.\n");
155         return -1;
156     }
157
158     p_schro_params->format->frame_rate_numerator   = avccontext->time_base.den;
159     p_schro_params->format->frame_rate_denominator = avccontext->time_base.num;
160
161     p_schro_params->frame_size = avpicture_get_size(avccontext->pix_fmt,
162                                                     avccontext->width,
163                                                     avccontext->height);
164
165     avccontext->coded_frame = &p_schro_params->picture;
166
167     if (!avccontext->gop_size) {
168         schro_encoder_setting_set_double(p_schro_params->encoder,
169                                          "gop_structure",
170                                          SCHRO_ENCODER_GOP_INTRA_ONLY);
171
172         if (avccontext->coder_type == FF_CODER_TYPE_VLC)
173             schro_encoder_setting_set_double(p_schro_params->encoder,
174                                              "enable_noarith", 1);
175     } else {
176         schro_encoder_setting_set_double(p_schro_params->encoder,
177                                          "au_distance", avccontext->gop_size);
178         avccontext->has_b_frames = 1;
179         p_schro_params->dts = -1;
180     }
181
182     /* FIXME - Need to handle SCHRO_ENCODER_RATE_CONTROL_LOW_DELAY. */
183     if (avccontext->flags & CODEC_FLAG_QSCALE) {
184         if (!avccontext->global_quality) {
185             /* lossless coding */
186             schro_encoder_setting_set_double(p_schro_params->encoder,
187                                              "rate_control",
188                                              SCHRO_ENCODER_RATE_CONTROL_LOSSLESS);
189         } else {
190             int quality;
191             schro_encoder_setting_set_double(p_schro_params->encoder,
192                                              "rate_control",
193                                              SCHRO_ENCODER_RATE_CONTROL_CONSTANT_QUALITY);
194
195             quality = avccontext->global_quality / FF_QP2LAMBDA;
196             if (quality > 10)
197                 quality = 10;
198             schro_encoder_setting_set_double(p_schro_params->encoder,
199                                              "quality", quality);
200         }
201     } else {
202         schro_encoder_setting_set_double(p_schro_params->encoder,
203                                          "rate_control",
204                                          SCHRO_ENCODER_RATE_CONTROL_CONSTANT_BITRATE);
205
206         schro_encoder_setting_set_double(p_schro_params->encoder,
207                                          "bitrate",
208                                          avccontext->bit_rate);
209
210     }
211
212     if (avccontext->flags & CODEC_FLAG_INTERLACED_ME)
213         /* All material can be coded as interlaced or progressive
214            irrespective of the type of source material. */
215         schro_encoder_setting_set_double(p_schro_params->encoder,
216                                          "interlaced_coding", 1);
217
218     schro_encoder_setting_set_double(p_schro_params->encoder, "open_gop",
219                                      !(avccontext->flags & CODEC_FLAG_CLOSED_GOP));
220
221     /* FIXME: Signal range hardcoded to 8-bit data until both libschroedinger
222      * and libdirac support other bit-depth data. */
223     schro_video_format_set_std_signal_range(p_schro_params->format,
224                                             SCHRO_SIGNAL_RANGE_8BIT_VIDEO);
225
226     /* Set the encoder format. */
227     schro_encoder_set_video_format(p_schro_params->encoder,
228                                    p_schro_params->format);
229
230     /* Set the debug level. */
231     schro_debug_set_level(avccontext->debug);
232
233     schro_encoder_start(p_schro_params->encoder);
234
235     /* Initialize the encoded frame queue. */
236     ff_schro_queue_init(&p_schro_params->enc_frame_queue);
237     return 0;
238 }
239
240 static SchroFrame *libschroedinger_frame_from_data(AVCodecContext *avccontext,
241                                                    const AVFrame *frame)
242 {
243     SchroEncoderParams *p_schro_params = avccontext->priv_data;
244     SchroFrame *in_frame;
245     /* Input line size may differ from what the codec supports. Especially
246      * when transcoding from one format to another. So use avpicture_layout
247      * to copy the frame. */
248     in_frame = ff_create_schro_frame(avccontext, p_schro_params->frame_format);
249
250     if (in_frame)
251         avpicture_layout((const AVPicture *)frame, avccontext->pix_fmt,
252                           avccontext->width, avccontext->height,
253                           in_frame->components[0].data,
254                           p_schro_params->frame_size);
255
256     return in_frame;
257 }
258
259 static void libschroedinger_free_frame(void *data)
260 {
261     FFSchroEncodedFrame *enc_frame = data;
262
263     av_freep(&enc_frame->p_encbuf);
264     av_free(enc_frame);
265 }
266
267 static int libschroedinger_encode_frame(AVCodecContext *avccontext, AVPacket *pkt,
268                                         const AVFrame *frame, int *got_packet)
269 {
270     int enc_size = 0;
271     SchroEncoderParams *p_schro_params = avccontext->priv_data;
272     SchroEncoder *encoder = p_schro_params->encoder;
273     struct FFSchroEncodedFrame *p_frame_output = NULL;
274     int go = 1;
275     SchroBuffer *enc_buf;
276     int presentation_frame;
277     int parse_code;
278     int last_frame_in_sequence = 0;
279     int pkt_size, ret;
280
281     if (!frame) {
282         /* Push end of sequence if not already signalled. */
283         if (!p_schro_params->eos_signalled) {
284             schro_encoder_end_of_stream(encoder);
285             p_schro_params->eos_signalled = 1;
286         }
287     } else {
288         /* Allocate frame data to schro input buffer. */
289         SchroFrame *in_frame = libschroedinger_frame_from_data(avccontext,
290                                                                frame);
291         /* Load next frame. */
292         schro_encoder_push_frame(encoder, in_frame);
293     }
294
295     if (p_schro_params->eos_pulled)
296         go = 0;
297
298     /* Now check to see if we have any output from the encoder. */
299     while (go) {
300         SchroStateEnum state;
301         state = schro_encoder_wait(encoder);
302         switch (state) {
303         case SCHRO_STATE_HAVE_BUFFER:
304         case SCHRO_STATE_END_OF_STREAM:
305             enc_buf = schro_encoder_pull(encoder, &presentation_frame);
306             av_assert0(enc_buf->length > 0);
307             parse_code = enc_buf->data[4];
308
309             /* All non-frame data is prepended to actual frame data to
310              * be able to set the pts correctly. So we don't write data
311              * to the frame output queue until we actually have a frame
312              */
313             p_schro_params->enc_buf = av_realloc(p_schro_params->enc_buf,
314                                                  p_schro_params->enc_buf_size + enc_buf->length);
315
316             memcpy(p_schro_params->enc_buf + p_schro_params->enc_buf_size,
317                    enc_buf->data, enc_buf->length);
318             p_schro_params->enc_buf_size += enc_buf->length;
319
320
321             if (state == SCHRO_STATE_END_OF_STREAM) {
322                 p_schro_params->eos_pulled = 1;
323                 go = 0;
324             }
325
326             if (!SCHRO_PARSE_CODE_IS_PICTURE(parse_code)) {
327                 schro_buffer_unref(enc_buf);
328                 break;
329             }
330
331             /* Create output frame. */
332             p_frame_output = av_mallocz(sizeof(FFSchroEncodedFrame));
333             /* Set output data. */
334             p_frame_output->size     = p_schro_params->enc_buf_size;
335             p_frame_output->p_encbuf = p_schro_params->enc_buf;
336             if (SCHRO_PARSE_CODE_IS_INTRA(parse_code) &&
337                 SCHRO_PARSE_CODE_IS_REFERENCE(parse_code))
338                 p_frame_output->key_frame = 1;
339
340             /* Parse the coded frame number from the bitstream. Bytes 14
341              * through 17 represesent the frame number. */
342             p_frame_output->frame_num = AV_RB32(enc_buf->data + 13);
343
344             ff_schro_queue_push_back(&p_schro_params->enc_frame_queue,
345                                      p_frame_output);
346             p_schro_params->enc_buf_size = 0;
347             p_schro_params->enc_buf      = NULL;
348
349             schro_buffer_unref(enc_buf);
350
351             break;
352
353         case SCHRO_STATE_NEED_FRAME:
354             go = 0;
355             break;
356
357         case SCHRO_STATE_AGAIN:
358             break;
359
360         default:
361             av_log(avccontext, AV_LOG_ERROR, "Unknown Schro Encoder state\n");
362             return -1;
363         }
364     }
365
366     /* Copy 'next' frame in queue. */
367
368     if (p_schro_params->enc_frame_queue.size == 1 &&
369         p_schro_params->eos_pulled)
370         last_frame_in_sequence = 1;
371
372     p_frame_output = ff_schro_queue_pop(&p_schro_params->enc_frame_queue);
373
374     if (!p_frame_output)
375         return 0;
376
377     pkt_size = p_frame_output->size;
378     if (last_frame_in_sequence && p_schro_params->enc_buf_size > 0)
379         pkt_size += p_schro_params->enc_buf_size;
380     if ((ret = ff_alloc_packet2(avccontext, pkt, pkt_size)) < 0)
381         goto error;
382
383     memcpy(pkt->data, p_frame_output->p_encbuf, p_frame_output->size);
384     avccontext->coded_frame->key_frame = p_frame_output->key_frame;
385     /* Use the frame number of the encoded frame as the pts. It is OK to
386      * do so since Dirac is a constant frame rate codec. It expects input
387      * to be of constant frame rate. */
388     pkt->pts =
389     avccontext->coded_frame->pts = p_frame_output->frame_num;
390     pkt->dts = p_schro_params->dts++;
391     enc_size = p_frame_output->size;
392
393     /* Append the end of sequence information to the last frame in the
394      * sequence. */
395     if (last_frame_in_sequence && p_schro_params->enc_buf_size > 0) {
396         memcpy(pkt->data + enc_size, p_schro_params->enc_buf,
397                p_schro_params->enc_buf_size);
398         enc_size += p_schro_params->enc_buf_size;
399         av_freep(&p_schro_params->enc_buf);
400         p_schro_params->enc_buf_size = 0;
401     }
402
403     if (p_frame_output->key_frame)
404         pkt->flags |= AV_PKT_FLAG_KEY;
405     *got_packet = 1;
406
407 error:
408     /* free frame */
409     libschroedinger_free_frame(p_frame_output);
410     return ret;
411 }
412
413
414 static int libschroedinger_encode_close(AVCodecContext *avccontext)
415 {
416     SchroEncoderParams *p_schro_params = avccontext->priv_data;
417
418     /* Close the encoder. */
419     schro_encoder_free(p_schro_params->encoder);
420
421     /* Free data in the output frame queue. */
422     ff_schro_queue_free(&p_schro_params->enc_frame_queue,
423                         libschroedinger_free_frame);
424
425
426     /* Free the encoder buffer. */
427     if (p_schro_params->enc_buf_size)
428         av_freep(&p_schro_params->enc_buf);
429
430     /* Free the video format structure. */
431     av_freep(&p_schro_params->format);
432
433     return 0;
434 }
435
436
437 AVCodec ff_libschroedinger_encoder = {
438     .name           = "libschroedinger",
439     .type           = AVMEDIA_TYPE_VIDEO,
440     .id             = AV_CODEC_ID_DIRAC,
441     .priv_data_size = sizeof(SchroEncoderParams),
442     .init           = libschroedinger_encode_init,
443     .encode2        = libschroedinger_encode_frame,
444     .close          = libschroedinger_encode_close,
445     .capabilities   = CODEC_CAP_DELAY,
446     .pix_fmts       = (const enum PixelFormat[]){
447         PIX_FMT_YUV420P, PIX_FMT_YUV422P, PIX_FMT_YUV444P, PIX_FMT_NONE
448     },
449     .long_name      = NULL_IF_CONFIG_SMALL("libschroedinger Dirac 2.2"),
450 };