utvideoenc: Add support for the new BT.709 FourCCs for YCbCr
[libav.git] / libavcodec / utvideoenc.c
1 /*
2 * Ut Video encoder
3 * Copyright (c) 2012 Jan Ekström
4 *
5 * This file is part of Libav.
6 *
7 * Libav 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 * Libav 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 Libav; 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 * Ut Video encoder
25 */
26
27 #include "libavutil/imgutils.h"
28 #include "libavutil/intreadwrite.h"
29 #include "avcodec.h"
30 #include "internal.h"
31 #include "bytestream.h"
32 #include "put_bits.h"
33 #include "dsputil.h"
34 #include "mathops.h"
35 #include "utvideo.h"
36 #include "huffman.h"
37
38 /* Compare huffentry symbols */
39 static int huff_cmp_sym(const void *a, const void *b)
40 {
41 const HuffEntry *aa = a, *bb = b;
42 return aa->sym - bb->sym;
43 }
44
45 static av_cold int utvideo_encode_close(AVCodecContext *avctx)
46 {
47 UtvideoContext *c = avctx->priv_data;
48 int i;
49
50 av_freep(&avctx->coded_frame);
51 av_freep(&c->slice_bits);
52 for (i = 0; i < 4; i++)
53 av_freep(&c->slice_buffer[i]);
54
55 return 0;
56 }
57
58 static av_cold int utvideo_encode_init(AVCodecContext *avctx)
59 {
60 UtvideoContext *c = avctx->priv_data;
61 int i;
62 uint32_t original_format;
63
64 c->avctx = avctx;
65 c->frame_info_size = 4;
66 c->slice_stride = FFALIGN(avctx->width, 32);
67
68 switch (avctx->pix_fmt) {
69 case AV_PIX_FMT_RGB24:
70 c->planes = 3;
71 avctx->codec_tag = MKTAG('U', 'L', 'R', 'G');
72 original_format = UTVIDEO_RGB;
73 break;
74 case AV_PIX_FMT_RGBA:
75 c->planes = 4;
76 avctx->codec_tag = MKTAG('U', 'L', 'R', 'A');
77 original_format = UTVIDEO_RGBA;
78 break;
79 case AV_PIX_FMT_YUV420P:
80 if (avctx->width & 1 || avctx->height & 1) {
81 av_log(avctx, AV_LOG_ERROR,
82 "4:2:0 video requires even width and height.\n");
83 return AVERROR_INVALIDDATA;
84 }
85 c->planes = 3;
86 if (avctx->colorspace == AVCOL_SPC_BT709)
87 avctx->codec_tag = MKTAG('U', 'L', 'H', '0');
88 else
89 avctx->codec_tag = MKTAG('U', 'L', 'Y', '0');
90 original_format = UTVIDEO_420;
91 break;
92 case AV_PIX_FMT_YUV422P:
93 if (avctx->width & 1) {
94 av_log(avctx, AV_LOG_ERROR,
95 "4:2:2 video requires even width.\n");
96 return AVERROR_INVALIDDATA;
97 }
98 c->planes = 3;
99 if (avctx->colorspace == AVCOL_SPC_BT709)
100 avctx->codec_tag = MKTAG('U', 'L', 'H', '2');
101 else
102 avctx->codec_tag = MKTAG('U', 'L', 'Y', '2');
103 original_format = UTVIDEO_422;
104 break;
105 default:
106 av_log(avctx, AV_LOG_ERROR, "Unknown pixel format: %d\n",
107 avctx->pix_fmt);
108 return AVERROR_INVALIDDATA;
109 }
110
111 ff_dsputil_init(&c->dsp, avctx);
112
113 /* Check the prediction method, and error out if unsupported */
114 if (avctx->prediction_method < 0 || avctx->prediction_method > 4) {
115 av_log(avctx, AV_LOG_WARNING,
116 "Prediction method %d is not supported in Ut Video.\n",
117 avctx->prediction_method);
118 return AVERROR_OPTION_NOT_FOUND;
119 }
120
121 if (avctx->prediction_method == FF_PRED_PLANE) {
122 av_log(avctx, AV_LOG_ERROR,
123 "Plane prediction is not supported in Ut Video.\n");
124 return AVERROR_OPTION_NOT_FOUND;
125 }
126
127 /* Convert from libavcodec prediction type to Ut Video's */
128 c->frame_pred = ff_ut_pred_order[avctx->prediction_method];
129
130 if (c->frame_pred == PRED_GRADIENT) {
131 av_log(avctx, AV_LOG_ERROR, "Gradient prediction is not supported.\n");
132 return AVERROR_OPTION_NOT_FOUND;
133 }
134
135 avctx->coded_frame = av_frame_alloc();
136
137 if (!avctx->coded_frame) {
138 av_log(avctx, AV_LOG_ERROR, "Could not allocate frame.\n");
139 utvideo_encode_close(avctx);
140 return AVERROR(ENOMEM);
141 }
142
143 /* extradata size is 4 * 32bit */
144 avctx->extradata_size = 16;
145
146 avctx->extradata = av_mallocz(avctx->extradata_size +
147 FF_INPUT_BUFFER_PADDING_SIZE);
148
149 if (!avctx->extradata) {
150 av_log(avctx, AV_LOG_ERROR, "Could not allocate extradata.\n");
151 utvideo_encode_close(avctx);
152 return AVERROR(ENOMEM);
153 }
154
155 for (i = 0; i < c->planes; i++) {
156 c->slice_buffer[i] = av_malloc(c->slice_stride * (avctx->height + 2) +
157 FF_INPUT_BUFFER_PADDING_SIZE);
158 if (!c->slice_buffer[i]) {
159 av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer 1.\n");
160 utvideo_encode_close(avctx);
161 return AVERROR(ENOMEM);
162 }
163 }
164
165 /*
166 * Set the version of the encoder.
167 * Last byte is "implementation ID", which is
168 * obtained from the creator of the format.
169 * Libavcodec has been assigned with the ID 0xF0.
170 */
171 AV_WB32(avctx->extradata, MKTAG(1, 0, 0, 0xF0));
172
173 /*
174 * Set the "original format"
175 * Not used for anything during decoding.
176 */
177 AV_WL32(avctx->extradata + 4, original_format);
178
179 /* Write 4 as the 'frame info size' */
180 AV_WL32(avctx->extradata + 8, c->frame_info_size);
181
182 /*
183 * Set how many slices are going to be used.
184 * Set one slice for now.
185 */
186 c->slices = 1;
187
188 /* Set compression mode */
189 c->compression = COMP_HUFF;
190
191 /*
192 * Set the encoding flags:
193 * - Slice count minus 1
194 * - Interlaced encoding mode flag, set to zero for now.
195 * - Compression mode (none/huff)
196 * And write the flags.
197 */
198 c->flags = (c->slices - 1) << 24;
199 c->flags |= 0 << 11; // bit field to signal interlaced encoding mode
200 c->flags |= c->compression;
201
202 AV_WL32(avctx->extradata + 12, c->flags);
203
204 return 0;
205 }
206
207 static void mangle_rgb_planes(uint8_t *dst[4], int dst_stride, uint8_t *src,
208 int step, int stride, int width, int height)
209 {
210 int i, j;
211 int k = 2 * dst_stride;
212 unsigned int g;
213
214 for (j = 0; j < height; j++) {
215 if (step == 3) {
216 for (i = 0; i < width * step; i += step) {
217 g = src[i + 1];
218 dst[0][k] = g;
219 g += 0x80;
220 dst[1][k] = src[i + 2] - g;
221 dst[2][k] = src[i + 0] - g;
222 k++;
223 }
224 } else {
225 for (i = 0; i < width * step; i += step) {
226 g = src[i + 1];
227 dst[0][k] = g;
228 g += 0x80;
229 dst[1][k] = src[i + 2] - g;
230 dst[2][k] = src[i + 0] - g;
231 dst[3][k] = src[i + 3];
232 k++;
233 }
234 }
235 k += dst_stride - width;
236 src += stride;
237 }
238 }
239
240 /* Write data to a plane with left prediction */
241 static void left_predict(uint8_t *src, uint8_t *dst, int stride,
242 int width, int height)
243 {
244 int i, j;
245 uint8_t prev;
246
247 prev = 0x80; /* Set the initial value */
248 for (j = 0; j < height; j++) {
249 for (i = 0; i < width; i++) {
250 *dst++ = src[i] - prev;
251 prev = src[i];
252 }
253 src += stride;
254 }
255 }
256
257 /* Write data to a plane with median prediction */
258 static void median_predict(UtvideoContext *c, uint8_t *src, uint8_t *dst, int stride,
259 int width, int height)
260 {
261 int i, j;
262 int A, B;
263 uint8_t prev;
264
265 /* First line uses left neighbour prediction */
266 prev = 0x80; /* Set the initial value */
267 for (i = 0; i < width; i++) {
268 *dst++ = src[i] - prev;
269 prev = src[i];
270 }
271
272 if (height == 1)
273 return;
274
275 src += stride;
276
277 /*
278 * Second line uses top prediction for the first sample,
279 * and median for the rest.
280 */
281 A = B = 0;
282
283 /* Rest of the coded part uses median prediction */
284 for (j = 1; j < height; j++) {
285 c->dsp.sub_hfyu_median_prediction(dst, src - stride, src, width, &A, &B);
286 dst += width;
287 src += stride;
288 }
289 }
290
291 /* Count the usage of values in a plane */
292 static void count_usage(uint8_t *src, int width,
293 int height, uint64_t *counts)
294 {
295 int i, j;
296
297 for (j = 0; j < height; j++) {
298 for (i = 0; i < width; i++) {
299 counts[src[i]]++;
300 }
301 src += width;
302 }
303 }
304
305 /* Calculate the actual huffman codes from the code lengths */
306 static void calculate_codes(HuffEntry *he)
307 {
308 int last, i;
309 uint32_t code;
310
311 qsort(he, 256, sizeof(*he), ff_ut_huff_cmp_len);
312
313 last = 255;
314 while (he[last].len == 255 && last)
315 last--;
316
317 code = 1;
318 for (i = last; i >= 0; i--) {
319 he[i].code = code >> (32 - he[i].len);
320 code += 0x80000000u >> (he[i].len - 1);
321 }
322
323 qsort(he, 256, sizeof(*he), huff_cmp_sym);
324 }
325
326 /* Write huffman bit codes to a memory block */
327 static int write_huff_codes(uint8_t *src, uint8_t *dst, int dst_size,
328 int width, int height, HuffEntry *he)
329 {
330 PutBitContext pb;
331 int i, j;
332 int count;
333
334 init_put_bits(&pb, dst, dst_size);
335
336 /* Write the codes */
337 for (j = 0; j < height; j++) {
338 for (i = 0; i < width; i++)
339 put_bits(&pb, he[src[i]].len, he[src[i]].code);
340
341 src += width;
342 }
343
344 /* Pad output to a 32bit boundary */
345 count = put_bits_count(&pb) & 0x1F;
346
347 if (count)
348 put_bits(&pb, 32 - count, 0);
349
350 /* Get the amount of bits written */
351 count = put_bits_count(&pb);
352
353 /* Flush the rest with zeroes */
354 flush_put_bits(&pb);
355
356 return count;
357 }
358
359 static int encode_plane(AVCodecContext *avctx, uint8_t *src,
360 uint8_t *dst, int stride,
361 int width, int height, PutByteContext *pb)
362 {
363 UtvideoContext *c = avctx->priv_data;
364 uint8_t lengths[256];
365 uint64_t counts[256] = { 0 };
366
367 HuffEntry he[256];
368
369 uint32_t offset = 0, slice_len = 0;
370 int i, sstart, send = 0;
371 int symbol;
372
373 /* Do prediction / make planes */
374 switch (c->frame_pred) {
375 case PRED_NONE:
376 for (i = 0; i < c->slices; i++) {
377 sstart = send;
378 send = height * (i + 1) / c->slices;
379 av_image_copy_plane(dst + sstart * width, width,
380 src + sstart * stride, stride,
381 width, send - sstart);
382 }
383 break;
384 case PRED_LEFT:
385 for (i = 0; i < c->slices; i++) {
386 sstart = send;
387 send = height * (i + 1) / c->slices;
388 left_predict(src + sstart * stride, dst + sstart * width,
389 stride, width, send - sstart);
390 }
391 break;
392 case PRED_MEDIAN:
393 for (i = 0; i < c->slices; i++) {
394 sstart = send;
395 send = height * (i + 1) / c->slices;
396 median_predict(c, src + sstart * stride, dst + sstart * width,
397 stride, width, send - sstart);
398 }
399 break;
400 default:
401 av_log(avctx, AV_LOG_ERROR, "Unknown prediction mode: %d\n",
402 c->frame_pred);
403 return AVERROR_OPTION_NOT_FOUND;
404 }
405
406 /* Count the usage of values */
407 count_usage(dst, width, height, counts);
408
409 /* Check for a special case where only one symbol was used */
410 for (symbol = 0; symbol < 256; symbol++) {
411 /* If non-zero count is found, see if it matches width * height */
412 if (counts[symbol]) {
413 /* Special case if only one symbol was used */
414 if (counts[symbol] == width * height) {
415 /*
416 * Write a zero for the single symbol
417 * used in the plane, else 0xFF.
418 */
419 for (i = 0; i < 256; i++) {
420 if (i == symbol)
421 bytestream2_put_byte(pb, 0);
422 else
423 bytestream2_put_byte(pb, 0xFF);
424 }
425
426 /* Write zeroes for lengths */
427 for (i = 0; i < c->slices; i++)
428 bytestream2_put_le32(pb, 0);
429
430 /* And that's all for that plane folks */
431 return 0;
432 }
433 break;
434 }
435 }
436
437 /* Calculate huffman lengths */
438 ff_huff_gen_len_table(lengths, counts);
439
440 /*
441 * Write the plane's header into the output packet:
442 * - huffman code lengths (256 bytes)
443 * - slice end offsets (gotten from the slice lengths)
444 */
445 for (i = 0; i < 256; i++) {
446 bytestream2_put_byte(pb, lengths[i]);
447
448 he[i].len = lengths[i];
449 he[i].sym = i;
450 }
451
452 /* Calculate the huffman codes themselves */
453 calculate_codes(he);
454
455 send = 0;
456 for (i = 0; i < c->slices; i++) {
457 sstart = send;
458 send = height * (i + 1) / c->slices;
459
460 /*
461 * Write the huffman codes to a buffer,
462 * get the offset in bits and convert to bytes.
463 */
464 offset += write_huff_codes(dst + sstart * width, c->slice_bits,
465 width * (send - sstart), width,
466 send - sstart, he) >> 3;
467
468 slice_len = offset - slice_len;
469
470 /* Byteswap the written huffman codes */
471 c->dsp.bswap_buf((uint32_t *) c->slice_bits,
472 (uint32_t *) c->slice_bits,
473 slice_len >> 2);
474
475 /* Write the offset to the stream */
476 bytestream2_put_le32(pb, offset);
477
478 /* Seek to the data part of the packet */
479 bytestream2_seek_p(pb, 4 * (c->slices - i - 1) +
480 offset - slice_len, SEEK_CUR);
481
482 /* Write the slices' data into the output packet */
483 bytestream2_put_buffer(pb, c->slice_bits, slice_len);
484
485 /* Seek back to the slice offsets */
486 bytestream2_seek_p(pb, -4 * (c->slices - i - 1) - offset,
487 SEEK_CUR);
488
489 slice_len = offset;
490 }
491
492 /* And at the end seek to the end of written slice(s) */
493 bytestream2_seek_p(pb, offset, SEEK_CUR);
494
495 return 0;
496 }
497
498 static int utvideo_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
499 const AVFrame *pic, int *got_packet)
500 {
501 UtvideoContext *c = avctx->priv_data;
502 PutByteContext pb;
503
504 uint32_t frame_info;
505
506 uint8_t *dst;
507
508 int width = avctx->width, height = avctx->height;
509 int i, ret = 0;
510
511 /* Allocate a new packet if needed, and set it to the pointer dst */
512 ret = ff_alloc_packet(pkt, (256 + 4 * c->slices + width * height) *
513 c->planes + 4);
514
515 if (ret < 0) {
516 av_log(avctx, AV_LOG_ERROR,
517 "Error allocating the output packet, or the provided packet "
518 "was too small.\n");
519 return ret;
520 }
521
522 dst = pkt->data;
523
524 bytestream2_init_writer(&pb, dst, pkt->size);
525
526 av_fast_malloc(&c->slice_bits, &c->slice_bits_size,
527 width * height + FF_INPUT_BUFFER_PADDING_SIZE);
528
529 if (!c->slice_bits) {
530 av_log(avctx, AV_LOG_ERROR, "Cannot allocate temporary buffer 2.\n");
531 return AVERROR(ENOMEM);
532 }
533
534 /* In case of RGB, mangle the planes to Ut Video's format */
535 if (avctx->pix_fmt == AV_PIX_FMT_RGBA || avctx->pix_fmt == AV_PIX_FMT_RGB24)
536 mangle_rgb_planes(c->slice_buffer, c->slice_stride, pic->data[0],
537 c->planes, pic->linesize[0], width, height);
538
539 /* Deal with the planes */
540 switch (avctx->pix_fmt) {
541 case AV_PIX_FMT_RGB24:
542 case AV_PIX_FMT_RGBA:
543 for (i = 0; i < c->planes; i++) {
544 ret = encode_plane(avctx, c->slice_buffer[i] + 2 * c->slice_stride,
545 c->slice_buffer[i], c->slice_stride,
546 width, height, &pb);
547
548 if (ret) {
549 av_log(avctx, AV_LOG_ERROR, "Error encoding plane %d.\n", i);
550 return ret;
551 }
552 }
553 break;
554 case AV_PIX_FMT_YUV422P:
555 for (i = 0; i < c->planes; i++) {
556 ret = encode_plane(avctx, pic->data[i], c->slice_buffer[0],
557 pic->linesize[i], width >> !!i, height, &pb);
558
559 if (ret) {
560 av_log(avctx, AV_LOG_ERROR, "Error encoding plane %d.\n", i);
561 return ret;
562 }
563 }
564 break;
565 case AV_PIX_FMT_YUV420P:
566 for (i = 0; i < c->planes; i++) {
567 ret = encode_plane(avctx, pic->data[i], c->slice_buffer[0],
568 pic->linesize[i], width >> !!i, height >> !!i,
569 &pb);
570
571 if (ret) {
572 av_log(avctx, AV_LOG_ERROR, "Error encoding plane %d.\n", i);
573 return ret;
574 }
575 }
576 break;
577 default:
578 av_log(avctx, AV_LOG_ERROR, "Unknown pixel format: %d\n",
579 avctx->pix_fmt);
580 return AVERROR_INVALIDDATA;
581 }
582
583 /*
584 * Write frame information (LE 32bit unsigned)
585 * into the output packet.
586 * Contains the prediction method.
587 */
588 frame_info = c->frame_pred << 8;
589 bytestream2_put_le32(&pb, frame_info);
590
591 /*
592 * At least currently Ut Video is IDR only.
593 * Set flags accordingly.
594 */
595 avctx->coded_frame->key_frame = 1;
596 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
597
598 pkt->size = bytestream2_tell_p(&pb);
599 pkt->flags |= AV_PKT_FLAG_KEY;
600
601 /* Packet should be done */
602 *got_packet = 1;
603
604 return 0;
605 }
606
607 AVCodec ff_utvideo_encoder = {
608 .name = "utvideo",
609 .long_name = NULL_IF_CONFIG_SMALL("Ut Video"),
610 .type = AVMEDIA_TYPE_VIDEO,
611 .id = AV_CODEC_ID_UTVIDEO,
612 .priv_data_size = sizeof(UtvideoContext),
613 .init = utvideo_encode_init,
614 .encode2 = utvideo_encode_frame,
615 .close = utvideo_encode_close,
616 .pix_fmts = (const enum AVPixelFormat[]) {
617 AV_PIX_FMT_RGB24, AV_PIX_FMT_RGBA, AV_PIX_FMT_YUV422P,
618 AV_PIX_FMT_YUV420P, AV_PIX_FMT_NONE
619 },
620 };