1 /*
2 * Copyright (c) 2002-2014 Michael Niedermayer <michaelni@gmx.at>
3 *
4 * see http://www.pcisys.net/~melanson/codecs/huffyuv.txt for a description of
5 * the algorithm used
6 *
7 * This file is part of FFmpeg.
8 *
9 * FFmpeg is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public
11 * License as published by the Free Software Foundation; either
12 * version 2.1 of the License, or (at your option) any later version.
13 *
14 * FFmpeg is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
18 *
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with FFmpeg; if not, write to the Free Software
21 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 *
23 * yuva, gray, 4:4:4, 4:1:1, 4:1:0 and >8 bit per sample support sponsored by NOA
24 */
25
26 /**
27 * @file
28 * huffyuv encoder
29 */
30
31 #include "avcodec.h"
32 #include "huffyuv.h"
33 #include "huffman.h"
34 #include "huffyuvencdsp.h"
35 #include "internal.h"
36 #include "lossless_videoencdsp.h"
37 #include "put_bits.h"
38 #include "libavutil/opt.h"
39 #include "libavutil/pixdesc.h"
40
diff_bytes(HYuvContext * s,uint8_t * dst,const uint8_t * src0,const uint8_t * src1,int w)41 static inline void diff_bytes(HYuvContext *s, uint8_t *dst,
42 const uint8_t *src0, const uint8_t *src1, int w)
43 {
44 if (s->bps <= 8) {
45 s->llvidencdsp.diff_bytes(dst, src0, src1, w);
46 } else {
47 s->hencdsp.diff_int16((uint16_t *)dst, (const uint16_t *)src0, (const uint16_t *)src1, s->n - 1, w);
48 }
49 }
50
sub_left_prediction(HYuvContext * s,uint8_t * dst,const uint8_t * src,int w,int left)51 static inline int sub_left_prediction(HYuvContext *s, uint8_t *dst,
52 const uint8_t *src, int w, int left)
53 {
54 int i;
55 int min_width = FFMIN(w, 32);
56
57 if (s->bps <= 8) {
58 for (i = 0; i < min_width; i++) { /* scalar loop before dsp call */
59 const int temp = src[i];
60 dst[i] = temp - left;
61 left = temp;
62 }
63 if (w < 32)
64 return left;
65 s->llvidencdsp.diff_bytes(dst + 32, src + 32, src + 31, w - 32);
66 return src[w-1];
67 } else {
68 const uint16_t *src16 = (const uint16_t *)src;
69 uint16_t *dst16 = ( uint16_t *)dst;
70 for (i = 0; i < min_width; i++) { /* scalar loop before dsp call */
71 const int temp = src16[i];
72 dst16[i] = temp - left;
73 left = temp;
74 }
75 if (w < 32)
76 return left;
77 s->hencdsp.diff_int16(dst16 + 32, src16 + 32, src16 + 31, s->n - 1, w - 32);
78 return src16[w-1];
79 }
80 }
81
sub_left_prediction_bgr32(HYuvContext * s,uint8_t * dst,const uint8_t * src,int w,int * red,int * green,int * blue,int * alpha)82 static inline void sub_left_prediction_bgr32(HYuvContext *s, uint8_t *dst,
83 const uint8_t *src, int w,
84 int *red, int *green, int *blue,
85 int *alpha)
86 {
87 int i;
88 int r, g, b, a;
89 int min_width = FFMIN(w, 8);
90 r = *red;
91 g = *green;
92 b = *blue;
93 a = *alpha;
94
95 for (i = 0; i < min_width; i++) {
96 const int rt = src[i * 4 + R];
97 const int gt = src[i * 4 + G];
98 const int bt = src[i * 4 + B];
99 const int at = src[i * 4 + A];
100 dst[i * 4 + R] = rt - r;
101 dst[i * 4 + G] = gt - g;
102 dst[i * 4 + B] = bt - b;
103 dst[i * 4 + A] = at - a;
104 r = rt;
105 g = gt;
106 b = bt;
107 a = at;
108 }
109
110 s->llvidencdsp.diff_bytes(dst + 32, src + 32, src + 32 - 4, w * 4 - 32);
111
112 *red = src[(w - 1) * 4 + R];
113 *green = src[(w - 1) * 4 + G];
114 *blue = src[(w - 1) * 4 + B];
115 *alpha = src[(w - 1) * 4 + A];
116 }
117
sub_left_prediction_rgb24(HYuvContext * s,uint8_t * dst,uint8_t * src,int w,int * red,int * green,int * blue)118 static inline void sub_left_prediction_rgb24(HYuvContext *s, uint8_t *dst,
119 uint8_t *src, int w,
120 int *red, int *green, int *blue)
121 {
122 int i;
123 int r, g, b;
124 r = *red;
125 g = *green;
126 b = *blue;
127 for (i = 0; i < FFMIN(w, 16); i++) {
128 const int rt = src[i * 3 + 0];
129 const int gt = src[i * 3 + 1];
130 const int bt = src[i * 3 + 2];
131 dst[i * 3 + 0] = rt - r;
132 dst[i * 3 + 1] = gt - g;
133 dst[i * 3 + 2] = bt - b;
134 r = rt;
135 g = gt;
136 b = bt;
137 }
138
139 s->llvidencdsp.diff_bytes(dst + 48, src + 48, src + 48 - 3, w * 3 - 48);
140
141 *red = src[(w - 1) * 3 + 0];
142 *green = src[(w - 1) * 3 + 1];
143 *blue = src[(w - 1) * 3 + 2];
144 }
145
sub_median_prediction(HYuvContext * s,uint8_t * dst,const uint8_t * src1,const uint8_t * src2,int w,int * left,int * left_top)146 static void sub_median_prediction(HYuvContext *s, uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top)
147 {
148 if (s->bps <= 8) {
149 s->llvidencdsp.sub_median_pred(dst, src1, src2, w , left, left_top);
150 } else {
151 s->hencdsp.sub_hfyu_median_pred_int16((uint16_t *)dst, (const uint16_t *)src1, (const uint16_t *)src2, s->n - 1, w , left, left_top);
152 }
153 }
154
store_table(HYuvContext * s,const uint8_t * len,uint8_t * buf)155 static int store_table(HYuvContext *s, const uint8_t *len, uint8_t *buf)
156 {
157 int i;
158 int index = 0;
159 int n = s->vlc_n;
160
161 for (i = 0; i < n;) {
162 int val = len[i];
163 int repeat = 0;
164
165 for (; i < n && len[i] == val && repeat < 255; i++)
166 repeat++;
167
168 av_assert0(val < 32 && val >0 && repeat < 256 && repeat>0);
169 if (repeat > 7) {
170 buf[index++] = val;
171 buf[index++] = repeat;
172 } else {
173 buf[index++] = val | (repeat << 5);
174 }
175 }
176
177 return index;
178 }
179
store_huffman_tables(HYuvContext * s,uint8_t * buf)180 static int store_huffman_tables(HYuvContext *s, uint8_t *buf)
181 {
182 int i, ret;
183 int size = 0;
184 int count = 3;
185
186 if (s->version > 2)
187 count = 1 + s->alpha + 2*s->chroma;
188
189 for (i = 0; i < count; i++) {
190 if ((ret = ff_huff_gen_len_table(s->len[i], s->stats[i], s->vlc_n, 0)) < 0)
191 return ret;
192
193 if (ff_huffyuv_generate_bits_table(s->bits[i], s->len[i], s->vlc_n) < 0) {
194 return -1;
195 }
196
197 size += store_table(s, s->len[i], buf + size);
198 }
199 return size;
200 }
201
encode_init(AVCodecContext * avctx)202 static av_cold int encode_init(AVCodecContext *avctx)
203 {
204 HYuvContext *s = avctx->priv_data;
205 int i, j;
206 int ret;
207 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(avctx->pix_fmt);
208
209 ff_huffyuv_common_init(avctx);
210 ff_huffyuvencdsp_init(&s->hencdsp, avctx);
211 ff_llvidencdsp_init(&s->llvidencdsp);
212
213 avctx->extradata = av_mallocz(3*MAX_N + 4);
214 if (s->flags&AV_CODEC_FLAG_PASS1) {
215 #define STATS_OUT_SIZE 21*MAX_N*3 + 4
216 avctx->stats_out = av_mallocz(STATS_OUT_SIZE); // 21*256*3(%llu ) + 3(\n) + 1(0) = 16132
217 if (!avctx->stats_out)
218 return AVERROR(ENOMEM);
219 }
220 s->version = 2;
221
222 if (!avctx->extradata)
223 return AVERROR(ENOMEM);
224
225 #if FF_API_CODED_FRAME
226 FF_DISABLE_DEPRECATION_WARNINGS
227 avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
228 avctx->coded_frame->key_frame = 1;
229 FF_ENABLE_DEPRECATION_WARNINGS
230 #endif
231 #if FF_API_PRIVATE_OPT
232 FF_DISABLE_DEPRECATION_WARNINGS
233 if (avctx->context_model == 1)
234 s->context = avctx->context_model;
235 FF_ENABLE_DEPRECATION_WARNINGS
236 #endif
237
238 s->bps = desc->comp[0].depth;
239 s->yuv = !(desc->flags & AV_PIX_FMT_FLAG_RGB) && desc->nb_components >= 2;
240 s->chroma = desc->nb_components > 2;
241 s->alpha = !!(desc->flags & AV_PIX_FMT_FLAG_ALPHA);
242 av_pix_fmt_get_chroma_sub_sample(avctx->pix_fmt,
243 &s->chroma_h_shift,
244 &s->chroma_v_shift);
245
246 switch (avctx->pix_fmt) {
247 case AV_PIX_FMT_YUV420P:
248 case AV_PIX_FMT_YUV422P:
249 if (s->width & 1) {
250 av_log(avctx, AV_LOG_ERROR, "Width must be even for this colorspace.\n");
251 return AVERROR(EINVAL);
252 }
253 s->bitstream_bpp = avctx->pix_fmt == AV_PIX_FMT_YUV420P ? 12 : 16;
254 break;
255 case AV_PIX_FMT_YUV444P:
256 case AV_PIX_FMT_YUV410P:
257 case AV_PIX_FMT_YUV411P:
258 case AV_PIX_FMT_YUV440P:
259 case AV_PIX_FMT_GBRP:
260 case AV_PIX_FMT_GBRP9:
261 case AV_PIX_FMT_GBRP10:
262 case AV_PIX_FMT_GBRP12:
263 case AV_PIX_FMT_GBRP14:
264 case AV_PIX_FMT_GBRP16:
265 case AV_PIX_FMT_GRAY8:
266 case AV_PIX_FMT_GRAY16:
267 case AV_PIX_FMT_YUVA444P:
268 case AV_PIX_FMT_YUVA420P:
269 case AV_PIX_FMT_YUVA422P:
270 case AV_PIX_FMT_GBRAP:
271 case AV_PIX_FMT_GRAY8A:
272 case AV_PIX_FMT_YUV420P9:
273 case AV_PIX_FMT_YUV420P10:
274 case AV_PIX_FMT_YUV420P12:
275 case AV_PIX_FMT_YUV420P14:
276 case AV_PIX_FMT_YUV420P16:
277 case AV_PIX_FMT_YUV422P9:
278 case AV_PIX_FMT_YUV422P10:
279 case AV_PIX_FMT_YUV422P12:
280 case AV_PIX_FMT_YUV422P14:
281 case AV_PIX_FMT_YUV422P16:
282 case AV_PIX_FMT_YUV444P9:
283 case AV_PIX_FMT_YUV444P10:
284 case AV_PIX_FMT_YUV444P12:
285 case AV_PIX_FMT_YUV444P14:
286 case AV_PIX_FMT_YUV444P16:
287 case AV_PIX_FMT_YUVA420P9:
288 case AV_PIX_FMT_YUVA420P10:
289 case AV_PIX_FMT_YUVA420P16:
290 case AV_PIX_FMT_YUVA422P9:
291 case AV_PIX_FMT_YUVA422P10:
292 case AV_PIX_FMT_YUVA422P16:
293 case AV_PIX_FMT_YUVA444P9:
294 case AV_PIX_FMT_YUVA444P10:
295 case AV_PIX_FMT_YUVA444P16:
296 s->version = 3;
297 break;
298 case AV_PIX_FMT_RGB32:
299 s->bitstream_bpp = 32;
300 break;
301 case AV_PIX_FMT_RGB24:
302 s->bitstream_bpp = 24;
303 break;
304 default:
305 av_log(avctx, AV_LOG_ERROR, "format not supported\n");
306 return AVERROR(EINVAL);
307 }
308 s->n = 1<<s->bps;
309 s->vlc_n = FFMIN(s->n, MAX_VLC_N);
310
311 avctx->bits_per_coded_sample = s->bitstream_bpp;
312 s->decorrelate = s->bitstream_bpp >= 24 && !s->yuv && !(desc->flags & AV_PIX_FMT_FLAG_PLANAR);
313 #if FF_API_PRIVATE_OPT
314 FF_DISABLE_DEPRECATION_WARNINGS
315 if (avctx->prediction_method)
316 s->predictor = avctx->prediction_method;
317 FF_ENABLE_DEPRECATION_WARNINGS
318 #endif
319 s->interlaced = avctx->flags & AV_CODEC_FLAG_INTERLACED_ME ? 1 : 0;
320 if (s->context) {
321 if (s->flags & (AV_CODEC_FLAG_PASS1 | AV_CODEC_FLAG_PASS2)) {
322 av_log(avctx, AV_LOG_ERROR,
323 "context=1 is not compatible with "
324 "2 pass huffyuv encoding\n");
325 return AVERROR(EINVAL);
326 }
327 }
328
329 if (avctx->codec->id == AV_CODEC_ID_HUFFYUV) {
330 if (avctx->pix_fmt == AV_PIX_FMT_YUV420P) {
331 av_log(avctx, AV_LOG_ERROR,
332 "Error: YV12 is not supported by huffyuv; use "
333 "vcodec=ffvhuff or format=422p\n");
334 return AVERROR(EINVAL);
335 }
336 #if FF_API_PRIVATE_OPT
337 if (s->context) {
338 av_log(avctx, AV_LOG_ERROR,
339 "Error: per-frame huffman tables are not supported "
340 "by huffyuv; use vcodec=ffvhuff\n");
341 return AVERROR(EINVAL);
342 }
343 if (s->version > 2) {
344 av_log(avctx, AV_LOG_ERROR,
345 "Error: ver>2 is not supported "
346 "by huffyuv; use vcodec=ffvhuff\n");
347 return AVERROR(EINVAL);
348 }
349 #endif
350 if (s->interlaced != ( s->height > 288 ))
351 av_log(avctx, AV_LOG_INFO,
352 "using huffyuv 2.2.0 or newer interlacing flag\n");
353 }
354
355 if (s->version > 3 && avctx->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL) {
356 av_log(avctx, AV_LOG_ERROR, "Ver > 3 is under development, files encoded with it may not be decodable with future versions!!!\n"
357 "Use vstrict=-2 / -strict -2 to use it anyway.\n");
358 return AVERROR(EINVAL);
359 }
360
361 if (s->bitstream_bpp >= 24 && s->predictor == MEDIAN && s->version <= 2) {
362 av_log(avctx, AV_LOG_ERROR,
363 "Error: RGB is incompatible with median predictor\n");
364 return AVERROR(EINVAL);
365 }
366
367 ((uint8_t*)avctx->extradata)[0] = s->predictor | (s->decorrelate << 6);
368 ((uint8_t*)avctx->extradata)[2] = s->interlaced ? 0x10 : 0x20;
369 if (s->context)
370 ((uint8_t*)avctx->extradata)[2] |= 0x40;
371 if (s->version < 3) {
372 ((uint8_t*)avctx->extradata)[1] = s->bitstream_bpp;
373 ((uint8_t*)avctx->extradata)[3] = 0;
374 } else {
375 ((uint8_t*)avctx->extradata)[1] = ((s->bps-1)<<4) | s->chroma_h_shift | (s->chroma_v_shift<<2);
376 if (s->chroma)
377 ((uint8_t*)avctx->extradata)[2] |= s->yuv ? 1 : 2;
378 if (s->alpha)
379 ((uint8_t*)avctx->extradata)[2] |= 4;
380 ((uint8_t*)avctx->extradata)[3] = 1;
381 }
382 s->avctx->extradata_size = 4;
383
384 if (avctx->stats_in) {
385 char *p = avctx->stats_in;
386
387 for (i = 0; i < 4; i++)
388 for (j = 0; j < s->vlc_n; j++)
389 s->stats[i][j] = 1;
390
391 for (;;) {
392 for (i = 0; i < 4; i++) {
393 char *next;
394
395 for (j = 0; j < s->vlc_n; j++) {
396 s->stats[i][j] += strtol(p, &next, 0);
397 if (next == p) return -1;
398 p = next;
399 }
400 }
401 if (p[0] == 0 || p[1] == 0 || p[2] == 0) break;
402 }
403 } else {
404 for (i = 0; i < 4; i++)
405 for (j = 0; j < s->vlc_n; j++) {
406 int d = FFMIN(j, s->vlc_n - j);
407
408 s->stats[i][j] = 100000000 / (d*d + 1);
409 }
410 }
411
412 ret = store_huffman_tables(s, s->avctx->extradata + s->avctx->extradata_size);
413 if (ret < 0)
414 return ret;
415 s->avctx->extradata_size += ret;
416
417 if (s->context) {
418 for (i = 0; i < 4; i++) {
419 int pels = s->width * s->height / (i ? 40 : 10);
420 for (j = 0; j < s->vlc_n; j++) {
421 int d = FFMIN(j, s->vlc_n - j);
422 s->stats[i][j] = pels/(d*d + 1);
423 }
424 }
425 } else {
426 for (i = 0; i < 4; i++)
427 for (j = 0; j < s->vlc_n; j++)
428 s->stats[i][j]= 0;
429 }
430
431 if (ff_huffyuv_alloc_temp(s)) {
432 ff_huffyuv_common_end(s);
433 return AVERROR(ENOMEM);
434 }
435
436 s->picture_number=0;
437
438 return 0;
439 }
encode_422_bitstream(HYuvContext * s,int offset,int count)440 static int encode_422_bitstream(HYuvContext *s, int offset, int count)
441 {
442 int i;
443 const uint8_t *y = s->temp[0] + offset;
444 const uint8_t *u = s->temp[1] + offset / 2;
445 const uint8_t *v = s->temp[2] + offset / 2;
446
447 if (s->pb.buf_end - s->pb.buf - (put_bits_count(&s->pb) >> 3) < 2 * 4 * count) {
448 av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
449 return -1;
450 }
451
452 #define LOAD4\
453 int y0 = y[2 * i];\
454 int y1 = y[2 * i + 1];\
455 int u0 = u[i];\
456 int v0 = v[i];
457
458 count /= 2;
459
460 if (s->flags & AV_CODEC_FLAG_PASS1) {
461 for(i = 0; i < count; i++) {
462 LOAD4;
463 s->stats[0][y0]++;
464 s->stats[1][u0]++;
465 s->stats[0][y1]++;
466 s->stats[2][v0]++;
467 }
468 }
469 if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
470 return 0;
471 if (s->context) {
472 for (i = 0; i < count; i++) {
473 LOAD4;
474 s->stats[0][y0]++;
475 put_bits(&s->pb, s->len[0][y0], s->bits[0][y0]);
476 s->stats[1][u0]++;
477 put_bits(&s->pb, s->len[1][u0], s->bits[1][u0]);
478 s->stats[0][y1]++;
479 put_bits(&s->pb, s->len[0][y1], s->bits[0][y1]);
480 s->stats[2][v0]++;
481 put_bits(&s->pb, s->len[2][v0], s->bits[2][v0]);
482 }
483 } else {
484 for(i = 0; i < count; i++) {
485 LOAD4;
486 put_bits(&s->pb, s->len[0][y0], s->bits[0][y0]);
487 put_bits(&s->pb, s->len[1][u0], s->bits[1][u0]);
488 put_bits(&s->pb, s->len[0][y1], s->bits[0][y1]);
489 put_bits(&s->pb, s->len[2][v0], s->bits[2][v0]);
490 }
491 }
492 return 0;
493 }
494
encode_plane_bitstream(HYuvContext * s,int width,int plane)495 static int encode_plane_bitstream(HYuvContext *s, int width, int plane)
496 {
497 int i, count = width/2;
498
499 if (s->pb.buf_end - s->pb.buf - (put_bits_count(&s->pb) >> 3) < count * s->bps / 2) {
500 av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
501 return -1;
502 }
503
504 #define LOADEND\
505 int y0 = s->temp[0][width-1];
506 #define LOADEND_14\
507 int y0 = s->temp16[0][width-1] & mask;
508 #define LOADEND_16\
509 int y0 = s->temp16[0][width-1];
510 #define STATEND\
511 s->stats[plane][y0]++;
512 #define STATEND_16\
513 s->stats[plane][y0>>2]++;
514 #define WRITEEND\
515 put_bits(&s->pb, s->len[plane][y0], s->bits[plane][y0]);
516 #define WRITEEND_16\
517 put_bits(&s->pb, s->len[plane][y0>>2], s->bits[plane][y0>>2]);\
518 put_bits(&s->pb, 2, y0&3);
519
520 #define LOAD2\
521 int y0 = s->temp[0][2 * i];\
522 int y1 = s->temp[0][2 * i + 1];
523 #define LOAD2_14\
524 int y0 = s->temp16[0][2 * i] & mask;\
525 int y1 = s->temp16[0][2 * i + 1] & mask;
526 #define LOAD2_16\
527 int y0 = s->temp16[0][2 * i];\
528 int y1 = s->temp16[0][2 * i + 1];
529 #define STAT2\
530 s->stats[plane][y0]++;\
531 s->stats[plane][y1]++;
532 #define STAT2_16\
533 s->stats[plane][y0>>2]++;\
534 s->stats[plane][y1>>2]++;
535 #define WRITE2\
536 put_bits(&s->pb, s->len[plane][y0], s->bits[plane][y0]);\
537 put_bits(&s->pb, s->len[plane][y1], s->bits[plane][y1]);
538 #define WRITE2_16\
539 put_bits(&s->pb, s->len[plane][y0>>2], s->bits[plane][y0>>2]);\
540 put_bits(&s->pb, 2, y0&3);\
541 put_bits(&s->pb, s->len[plane][y1>>2], s->bits[plane][y1>>2]);\
542 put_bits(&s->pb, 2, y1&3);
543
544 if (s->bps <= 8) {
545 if (s->flags & AV_CODEC_FLAG_PASS1) {
546 for (i = 0; i < count; i++) {
547 LOAD2;
548 STAT2;
549 }
550 if (width&1) {
551 LOADEND;
552 STATEND;
553 }
554 }
555 if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
556 return 0;
557
558 if (s->context) {
559 for (i = 0; i < count; i++) {
560 LOAD2;
561 STAT2;
562 WRITE2;
563 }
564 if (width&1) {
565 LOADEND;
566 STATEND;
567 WRITEEND;
568 }
569 } else {
570 for (i = 0; i < count; i++) {
571 LOAD2;
572 WRITE2;
573 }
574 if (width&1) {
575 LOADEND;
576 WRITEEND;
577 }
578 }
579 } else if (s->bps <= 14) {
580 int mask = s->n - 1;
581 if (s->flags & AV_CODEC_FLAG_PASS1) {
582 for (i = 0; i < count; i++) {
583 LOAD2_14;
584 STAT2;
585 }
586 if (width&1) {
587 LOADEND_14;
588 STATEND;
589 }
590 }
591 if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
592 return 0;
593
594 if (s->context) {
595 for (i = 0; i < count; i++) {
596 LOAD2_14;
597 STAT2;
598 WRITE2;
599 }
600 if (width&1) {
601 LOADEND_14;
602 STATEND;
603 WRITEEND;
604 }
605 } else {
606 for (i = 0; i < count; i++) {
607 LOAD2_14;
608 WRITE2;
609 }
610 if (width&1) {
611 LOADEND_14;
612 WRITEEND;
613 }
614 }
615 } else {
616 if (s->flags & AV_CODEC_FLAG_PASS1) {
617 for (i = 0; i < count; i++) {
618 LOAD2_16;
619 STAT2_16;
620 }
621 if (width&1) {
622 LOADEND_16;
623 STATEND_16;
624 }
625 }
626 if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
627 return 0;
628
629 if (s->context) {
630 for (i = 0; i < count; i++) {
631 LOAD2_16;
632 STAT2_16;
633 WRITE2_16;
634 }
635 if (width&1) {
636 LOADEND_16;
637 STATEND_16;
638 WRITEEND_16;
639 }
640 } else {
641 for (i = 0; i < count; i++) {
642 LOAD2_16;
643 WRITE2_16;
644 }
645 if (width&1) {
646 LOADEND_16;
647 WRITEEND_16;
648 }
649 }
650 }
651 #undef LOAD2
652 #undef STAT2
653 #undef WRITE2
654 return 0;
655 }
656
encode_gray_bitstream(HYuvContext * s,int count)657 static int encode_gray_bitstream(HYuvContext *s, int count)
658 {
659 int i;
660
661 if (s->pb.buf_end - s->pb.buf - (put_bits_count(&s->pb) >> 3) < 4 * count) {
662 av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
663 return -1;
664 }
665
666 #define LOAD2\
667 int y0 = s->temp[0][2 * i];\
668 int y1 = s->temp[0][2 * i + 1];
669 #define STAT2\
670 s->stats[0][y0]++;\
671 s->stats[0][y1]++;
672 #define WRITE2\
673 put_bits(&s->pb, s->len[0][y0], s->bits[0][y0]);\
674 put_bits(&s->pb, s->len[0][y1], s->bits[0][y1]);
675
676 count /= 2;
677
678 if (s->flags & AV_CODEC_FLAG_PASS1) {
679 for (i = 0; i < count; i++) {
680 LOAD2;
681 STAT2;
682 }
683 }
684 if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
685 return 0;
686
687 if (s->context) {
688 for (i = 0; i < count; i++) {
689 LOAD2;
690 STAT2;
691 WRITE2;
692 }
693 } else {
694 for (i = 0; i < count; i++) {
695 LOAD2;
696 WRITE2;
697 }
698 }
699 return 0;
700 }
701
encode_bgra_bitstream(HYuvContext * s,int count,int planes)702 static inline int encode_bgra_bitstream(HYuvContext *s, int count, int planes)
703 {
704 int i;
705
706 if (s->pb.buf_end - s->pb.buf - (put_bits_count(&s->pb) >> 3) <
707 4 * planes * count) {
708 av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
709 return -1;
710 }
711
712 #define LOAD_GBRA \
713 int g = s->temp[0][planes == 3 ? 3 * i + 1 : 4 * i + G]; \
714 int b =(s->temp[0][planes == 3 ? 3 * i + 2 : 4 * i + B] - g) & 0xFF;\
715 int r =(s->temp[0][planes == 3 ? 3 * i + 0 : 4 * i + R] - g) & 0xFF;\
716 int a = s->temp[0][planes * i + A];
717
718 #define STAT_BGRA \
719 s->stats[0][b]++; \
720 s->stats[1][g]++; \
721 s->stats[2][r]++; \
722 if (planes == 4) \
723 s->stats[2][a]++;
724
725 #define WRITE_GBRA \
726 put_bits(&s->pb, s->len[1][g], s->bits[1][g]); \
727 put_bits(&s->pb, s->len[0][b], s->bits[0][b]); \
728 put_bits(&s->pb, s->len[2][r], s->bits[2][r]); \
729 if (planes == 4) \
730 put_bits(&s->pb, s->len[2][a], s->bits[2][a]);
731
732 if ((s->flags & AV_CODEC_FLAG_PASS1) &&
733 (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)) {
734 for (i = 0; i < count; i++) {
735 LOAD_GBRA;
736 STAT_BGRA;
737 }
738 } else if (s->context || (s->flags & AV_CODEC_FLAG_PASS1)) {
739 for (i = 0; i < count; i++) {
740 LOAD_GBRA;
741 STAT_BGRA;
742 WRITE_GBRA;
743 }
744 } else {
745 for (i = 0; i < count; i++) {
746 LOAD_GBRA;
747 WRITE_GBRA;
748 }
749 }
750 return 0;
751 }
752
encode_frame(AVCodecContext * avctx,AVPacket * pkt,const AVFrame * pict,int * got_packet)753 static int encode_frame(AVCodecContext *avctx, AVPacket *pkt,
754 const AVFrame *pict, int *got_packet)
755 {
756 HYuvContext *s = avctx->priv_data;
757 const int width = s->width;
758 const int width2 = s->width>>1;
759 const int height = s->height;
760 const int fake_ystride = s->interlaced ? pict->linesize[0]*2 : pict->linesize[0];
761 const int fake_ustride = s->interlaced ? pict->linesize[1]*2 : pict->linesize[1];
762 const int fake_vstride = s->interlaced ? pict->linesize[2]*2 : pict->linesize[2];
763 const AVFrame * const p = pict;
764 int i, j, size = 0, ret;
765
766 if ((ret = ff_alloc_packet2(avctx, pkt, width * height * 3 * 4 + AV_INPUT_BUFFER_MIN_SIZE, 0)) < 0)
767 return ret;
768
769 if (s->context) {
770 size = store_huffman_tables(s, pkt->data);
771 if (size < 0)
772 return size;
773
774 for (i = 0; i < 4; i++)
775 for (j = 0; j < s->vlc_n; j++)
776 s->stats[i][j] >>= 1;
777 }
778
779 init_put_bits(&s->pb, pkt->data + size, pkt->size - size);
780
781 if (avctx->pix_fmt == AV_PIX_FMT_YUV422P ||
782 avctx->pix_fmt == AV_PIX_FMT_YUV420P) {
783 int lefty, leftu, leftv, y, cy;
784
785 put_bits(&s->pb, 8, leftv = p->data[2][0]);
786 put_bits(&s->pb, 8, lefty = p->data[0][1]);
787 put_bits(&s->pb, 8, leftu = p->data[1][0]);
788 put_bits(&s->pb, 8, p->data[0][0]);
789
790 lefty = sub_left_prediction(s, s->temp[0], p->data[0], width , 0);
791 leftu = sub_left_prediction(s, s->temp[1], p->data[1], width2, 0);
792 leftv = sub_left_prediction(s, s->temp[2], p->data[2], width2, 0);
793
794 encode_422_bitstream(s, 2, width-2);
795
796 if (s->predictor==MEDIAN) {
797 int lefttopy, lefttopu, lefttopv;
798 cy = y = 1;
799 if (s->interlaced) {
800 lefty = sub_left_prediction(s, s->temp[0], p->data[0] + p->linesize[0], width , lefty);
801 leftu = sub_left_prediction(s, s->temp[1], p->data[1] + p->linesize[1], width2, leftu);
802 leftv = sub_left_prediction(s, s->temp[2], p->data[2] + p->linesize[2], width2, leftv);
803
804 encode_422_bitstream(s, 0, width);
805 y++; cy++;
806 }
807
808 lefty = sub_left_prediction(s, s->temp[0], p->data[0] + fake_ystride, 4, lefty);
809 leftu = sub_left_prediction(s, s->temp[1], p->data[1] + fake_ustride, 2, leftu);
810 leftv = sub_left_prediction(s, s->temp[2], p->data[2] + fake_vstride, 2, leftv);
811
812 encode_422_bitstream(s, 0, 4);
813
814 lefttopy = p->data[0][3];
815 lefttopu = p->data[1][1];
816 lefttopv = p->data[2][1];
817 s->llvidencdsp.sub_median_pred(s->temp[0], p->data[0] + 4, p->data[0] + fake_ystride + 4, width - 4, &lefty, &lefttopy);
818 s->llvidencdsp.sub_median_pred(s->temp[1], p->data[1] + 2, p->data[1] + fake_ustride + 2, width2 - 2, &leftu, &lefttopu);
819 s->llvidencdsp.sub_median_pred(s->temp[2], p->data[2] + 2, p->data[2] + fake_vstride + 2, width2 - 2, &leftv, &lefttopv);
820 encode_422_bitstream(s, 0, width - 4);
821 y++; cy++;
822
823 for (; y < height; y++,cy++) {
824 uint8_t *ydst, *udst, *vdst;
825
826 if (s->bitstream_bpp == 12) {
827 while (2 * cy > y) {
828 ydst = p->data[0] + p->linesize[0] * y;
829 s->llvidencdsp.sub_median_pred(s->temp[0], ydst - fake_ystride, ydst, width, &lefty, &lefttopy);
830 encode_gray_bitstream(s, width);
831 y++;
832 }
833 if (y >= height) break;
834 }
835 ydst = p->data[0] + p->linesize[0] * y;
836 udst = p->data[1] + p->linesize[1] * cy;
837 vdst = p->data[2] + p->linesize[2] * cy;
838
839 s->llvidencdsp.sub_median_pred(s->temp[0], ydst - fake_ystride, ydst, width, &lefty, &lefttopy);
840 s->llvidencdsp.sub_median_pred(s->temp[1], udst - fake_ustride, udst, width2, &leftu, &lefttopu);
841 s->llvidencdsp.sub_median_pred(s->temp[2], vdst - fake_vstride, vdst, width2, &leftv, &lefttopv);
842
843 encode_422_bitstream(s, 0, width);
844 }
845 } else {
846 for (cy = y = 1; y < height; y++, cy++) {
847 uint8_t *ydst, *udst, *vdst;
848
849 /* encode a luma only line & y++ */
850 if (s->bitstream_bpp == 12) {
851 ydst = p->data[0] + p->linesize[0] * y;
852
853 if (s->predictor == PLANE && s->interlaced < y) {
854 s->llvidencdsp.diff_bytes(s->temp[1], ydst, ydst - fake_ystride, width);
855
856 lefty = sub_left_prediction(s, s->temp[0], s->temp[1], width , lefty);
857 } else {
858 lefty = sub_left_prediction(s, s->temp[0], ydst, width , lefty);
859 }
860 encode_gray_bitstream(s, width);
861 y++;
862 if (y >= height) break;
863 }
864
865 ydst = p->data[0] + p->linesize[0] * y;
866 udst = p->data[1] + p->linesize[1] * cy;
867 vdst = p->data[2] + p->linesize[2] * cy;
868
869 if (s->predictor == PLANE && s->interlaced < cy) {
870 s->llvidencdsp.diff_bytes(s->temp[1], ydst, ydst - fake_ystride, width);
871 s->llvidencdsp.diff_bytes(s->temp[2], udst, udst - fake_ustride, width2);
872 s->llvidencdsp.diff_bytes(s->temp[2] + width2, vdst, vdst - fake_vstride, width2);
873
874 lefty = sub_left_prediction(s, s->temp[0], s->temp[1], width , lefty);
875 leftu = sub_left_prediction(s, s->temp[1], s->temp[2], width2, leftu);
876 leftv = sub_left_prediction(s, s->temp[2], s->temp[2] + width2, width2, leftv);
877 } else {
878 lefty = sub_left_prediction(s, s->temp[0], ydst, width , lefty);
879 leftu = sub_left_prediction(s, s->temp[1], udst, width2, leftu);
880 leftv = sub_left_prediction(s, s->temp[2], vdst, width2, leftv);
881 }
882
883 encode_422_bitstream(s, 0, width);
884 }
885 }
886 } else if(avctx->pix_fmt == AV_PIX_FMT_RGB32) {
887 uint8_t *data = p->data[0] + (height - 1) * p->linesize[0];
888 const int stride = -p->linesize[0];
889 const int fake_stride = -fake_ystride;
890 int y;
891 int leftr, leftg, leftb, lefta;
892
893 put_bits(&s->pb, 8, lefta = data[A]);
894 put_bits(&s->pb, 8, leftr = data[R]);
895 put_bits(&s->pb, 8, leftg = data[G]);
896 put_bits(&s->pb, 8, leftb = data[B]);
897
898 sub_left_prediction_bgr32(s, s->temp[0], data + 4, width - 1,
899 &leftr, &leftg, &leftb, &lefta);
900 encode_bgra_bitstream(s, width - 1, 4);
901
902 for (y = 1; y < s->height; y++) {
903 uint8_t *dst = data + y*stride;
904 if (s->predictor == PLANE && s->interlaced < y) {
905 s->llvidencdsp.diff_bytes(s->temp[1], dst, dst - fake_stride, width * 4);
906 sub_left_prediction_bgr32(s, s->temp[0], s->temp[1], width,
907 &leftr, &leftg, &leftb, &lefta);
908 } else {
909 sub_left_prediction_bgr32(s, s->temp[0], dst, width,
910 &leftr, &leftg, &leftb, &lefta);
911 }
912 encode_bgra_bitstream(s, width, 4);
913 }
914 } else if (avctx->pix_fmt == AV_PIX_FMT_RGB24) {
915 uint8_t *data = p->data[0] + (height - 1) * p->linesize[0];
916 const int stride = -p->linesize[0];
917 const int fake_stride = -fake_ystride;
918 int y;
919 int leftr, leftg, leftb;
920
921 put_bits(&s->pb, 8, leftr = data[0]);
922 put_bits(&s->pb, 8, leftg = data[1]);
923 put_bits(&s->pb, 8, leftb = data[2]);
924 put_bits(&s->pb, 8, 0);
925
926 sub_left_prediction_rgb24(s, s->temp[0], data + 3, width - 1,
927 &leftr, &leftg, &leftb);
928 encode_bgra_bitstream(s, width-1, 3);
929
930 for (y = 1; y < s->height; y++) {
931 uint8_t *dst = data + y * stride;
932 if (s->predictor == PLANE && s->interlaced < y) {
933 s->llvidencdsp.diff_bytes(s->temp[1], dst, dst - fake_stride,
934 width * 3);
935 sub_left_prediction_rgb24(s, s->temp[0], s->temp[1], width,
936 &leftr, &leftg, &leftb);
937 } else {
938 sub_left_prediction_rgb24(s, s->temp[0], dst, width,
939 &leftr, &leftg, &leftb);
940 }
941 encode_bgra_bitstream(s, width, 3);
942 }
943 } else if (s->version > 2) {
944 int plane;
945 for (plane = 0; plane < 1 + 2*s->chroma + s->alpha; plane++) {
946 int left, y;
947 int w = width;
948 int h = height;
949 int fake_stride = fake_ystride;
950
951 if (s->chroma && (plane == 1 || plane == 2)) {
952 w >>= s->chroma_h_shift;
953 h >>= s->chroma_v_shift;
954 fake_stride = plane == 1 ? fake_ustride : fake_vstride;
955 }
956
957 left = sub_left_prediction(s, s->temp[0], p->data[plane], w , 0);
958
959 encode_plane_bitstream(s, w, plane);
960
961 if (s->predictor==MEDIAN) {
962 int lefttop;
963 y = 1;
964 if (s->interlaced) {
965 left = sub_left_prediction(s, s->temp[0], p->data[plane] + p->linesize[plane], w , left);
966
967 encode_plane_bitstream(s, w, plane);
968 y++;
969 }
970
971 lefttop = p->data[plane][0];
972
973 for (; y < h; y++) {
974 uint8_t *dst = p->data[plane] + p->linesize[plane] * y;
975
976 sub_median_prediction(s, s->temp[0], dst - fake_stride, dst, w , &left, &lefttop);
977
978 encode_plane_bitstream(s, w, plane);
979 }
980 } else {
981 for (y = 1; y < h; y++) {
982 uint8_t *dst = p->data[plane] + p->linesize[plane] * y;
983
984 if (s->predictor == PLANE && s->interlaced < y) {
985 diff_bytes(s, s->temp[1], dst, dst - fake_stride, w);
986
987 left = sub_left_prediction(s, s->temp[0], s->temp[1], w , left);
988 } else {
989 left = sub_left_prediction(s, s->temp[0], dst, w , left);
990 }
991
992 encode_plane_bitstream(s, w, plane);
993 }
994 }
995 }
996 } else {
997 av_log(avctx, AV_LOG_ERROR, "Format not supported!\n");
998 }
999 emms_c();
1000
1001 size += (put_bits_count(&s->pb) + 31) / 8;
1002 put_bits(&s->pb, 16, 0);
1003 put_bits(&s->pb, 15, 0);
1004 size /= 4;
1005
1006 if ((s->flags & AV_CODEC_FLAG_PASS1) && (s->picture_number & 31) == 0) {
1007 int j;
1008 char *p = avctx->stats_out;
1009 char *end = p + STATS_OUT_SIZE;
1010 for (i = 0; i < 4; i++) {
1011 for (j = 0; j < s->vlc_n; j++) {
1012 snprintf(p, end-p, "%"PRIu64" ", s->stats[i][j]);
1013 p += strlen(p);
1014 s->stats[i][j]= 0;
1015 }
1016 snprintf(p, end-p, "\n");
1017 p++;
1018 if (end <= p)
1019 return AVERROR(ENOMEM);
1020 }
1021 } else if (avctx->stats_out)
1022 avctx->stats_out[0] = '\0';
1023 if (!(s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)) {
1024 flush_put_bits(&s->pb);
1025 s->bdsp.bswap_buf((uint32_t *) pkt->data, (uint32_t *) pkt->data, size);
1026 }
1027
1028 s->picture_number++;
1029
1030 pkt->size = size * 4;
1031 pkt->flags |= AV_PKT_FLAG_KEY;
1032 *got_packet = 1;
1033
1034 return 0;
1035 }
1036
encode_end(AVCodecContext * avctx)1037 static av_cold int encode_end(AVCodecContext *avctx)
1038 {
1039 HYuvContext *s = avctx->priv_data;
1040
1041 ff_huffyuv_common_end(s);
1042
1043 av_freep(&avctx->extradata);
1044 av_freep(&avctx->stats_out);
1045
1046 return 0;
1047 }
1048
1049 #define OFFSET(x) offsetof(HYuvContext, x)
1050 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
1051
1052 #define COMMON_OPTIONS \
1053 { "non_deterministic", "Allow multithreading for e.g. context=1 at the expense of determinism", \
1054 OFFSET(non_determ), AV_OPT_TYPE_BOOL, { .i64 = 1 }, \
1055 0, 1, VE }, \
1056 { "pred", "Prediction method", OFFSET(predictor), AV_OPT_TYPE_INT, { .i64 = LEFT }, LEFT, MEDIAN, VE, "pred" }, \
1057 { "left", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = LEFT }, INT_MIN, INT_MAX, VE, "pred" }, \
1058 { "plane", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = PLANE }, INT_MIN, INT_MAX, VE, "pred" }, \
1059 { "median", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MEDIAN }, INT_MIN, INT_MAX, VE, "pred" }, \
1060
1061 static const AVOption normal_options[] = {
1062 COMMON_OPTIONS
1063 { NULL },
1064 };
1065
1066 static const AVOption ff_options[] = {
1067 COMMON_OPTIONS
1068 { "context", "Set per-frame huffman tables", OFFSET(context), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE },
1069 { NULL },
1070 };
1071
1072 static const AVClass normal_class = {
1073 .class_name = "huffyuv",
1074 .item_name = av_default_item_name,
1075 .option = normal_options,
1076 .version = LIBAVUTIL_VERSION_INT,
1077 };
1078
1079 static const AVClass ff_class = {
1080 .class_name = "ffvhuff",
1081 .item_name = av_default_item_name,
1082 .option = ff_options,
1083 .version = LIBAVUTIL_VERSION_INT,
1084 };
1085
1086 AVCodec ff_huffyuv_encoder = {
1087 .name = "huffyuv",
1088 .long_name = NULL_IF_CONFIG_SMALL("Huffyuv / HuffYUV"),
1089 .type = AVMEDIA_TYPE_VIDEO,
1090 .id = AV_CODEC_ID_HUFFYUV,
1091 .priv_data_size = sizeof(HYuvContext),
1092 .init = encode_init,
1093 .encode2 = encode_frame,
1094 .close = encode_end,
1095 .capabilities = AV_CODEC_CAP_FRAME_THREADS | AV_CODEC_CAP_INTRA_ONLY,
1096 .priv_class = &normal_class,
1097 .pix_fmts = (const enum AVPixelFormat[]){
1098 AV_PIX_FMT_YUV422P, AV_PIX_FMT_RGB24,
1099 AV_PIX_FMT_RGB32, AV_PIX_FMT_NONE
1100 },
1101 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE |
1102 FF_CODEC_CAP_INIT_CLEANUP,
1103 };
1104
1105 #if CONFIG_FFVHUFF_ENCODER
1106 AVCodec ff_ffvhuff_encoder = {
1107 .name = "ffvhuff",
1108 .long_name = NULL_IF_CONFIG_SMALL("Huffyuv FFmpeg variant"),
1109 .type = AVMEDIA_TYPE_VIDEO,
1110 .id = AV_CODEC_ID_FFVHUFF,
1111 .priv_data_size = sizeof(HYuvContext),
1112 .init = encode_init,
1113 .encode2 = encode_frame,
1114 .close = encode_end,
1115 .capabilities = AV_CODEC_CAP_FRAME_THREADS | AV_CODEC_CAP_INTRA_ONLY,
1116 .priv_class = &ff_class,
1117 .pix_fmts = (const enum AVPixelFormat[]){
1118 AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV411P,
1119 AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUV440P,
1120 AV_PIX_FMT_GBRP,
1121 AV_PIX_FMT_GBRP9, AV_PIX_FMT_GBRP10, AV_PIX_FMT_GBRP12, AV_PIX_FMT_GBRP14,
1122 AV_PIX_FMT_GRAY8, AV_PIX_FMT_GRAY16,
1123 AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUVA444P,
1124 AV_PIX_FMT_GBRAP,
1125 AV_PIX_FMT_GRAY8A,
1126 AV_PIX_FMT_YUV420P9, AV_PIX_FMT_YUV420P10, AV_PIX_FMT_YUV420P12, AV_PIX_FMT_YUV420P14, AV_PIX_FMT_YUV420P16,
1127 AV_PIX_FMT_YUV422P9, AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUV422P12, AV_PIX_FMT_YUV422P14, AV_PIX_FMT_YUV422P16,
1128 AV_PIX_FMT_YUV444P9, AV_PIX_FMT_YUV444P10, AV_PIX_FMT_YUV444P12, AV_PIX_FMT_YUV444P14, AV_PIX_FMT_YUV444P16,
1129 AV_PIX_FMT_YUVA420P9, AV_PIX_FMT_YUVA420P10, AV_PIX_FMT_YUVA420P16,
1130 AV_PIX_FMT_YUVA422P9, AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_YUVA422P16,
1131 AV_PIX_FMT_YUVA444P9, AV_PIX_FMT_YUVA444P10, AV_PIX_FMT_YUVA444P16,
1132 AV_PIX_FMT_RGB24,
1133 AV_PIX_FMT_RGB32, AV_PIX_FMT_NONE
1134 },
1135 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE |
1136 FF_CODEC_CAP_INIT_CLEANUP,
1137 };
1138 #endif
1139