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_YUV420P9:
272 case AV_PIX_FMT_YUV420P10:
273 case AV_PIX_FMT_YUV420P12:
274 case AV_PIX_FMT_YUV420P14:
275 case AV_PIX_FMT_YUV420P16:
276 case AV_PIX_FMT_YUV422P9:
277 case AV_PIX_FMT_YUV422P10:
278 case AV_PIX_FMT_YUV422P12:
279 case AV_PIX_FMT_YUV422P14:
280 case AV_PIX_FMT_YUV422P16:
281 case AV_PIX_FMT_YUV444P9:
282 case AV_PIX_FMT_YUV444P10:
283 case AV_PIX_FMT_YUV444P12:
284 case AV_PIX_FMT_YUV444P14:
285 case AV_PIX_FMT_YUV444P16:
286 case AV_PIX_FMT_YUVA420P9:
287 case AV_PIX_FMT_YUVA420P10:
288 case AV_PIX_FMT_YUVA420P16:
289 case AV_PIX_FMT_YUVA422P9:
290 case AV_PIX_FMT_YUVA422P10:
291 case AV_PIX_FMT_YUVA422P16:
292 case AV_PIX_FMT_YUVA444P9:
293 case AV_PIX_FMT_YUVA444P10:
294 case AV_PIX_FMT_YUVA444P16:
295 s->version = 3;
296 break;
297 case AV_PIX_FMT_RGB32:
298 s->bitstream_bpp = 32;
299 break;
300 case AV_PIX_FMT_RGB24:
301 s->bitstream_bpp = 24;
302 break;
303 default:
304 av_log(avctx, AV_LOG_ERROR, "format not supported\n");
305 return AVERROR(EINVAL);
306 }
307 s->n = 1<<s->bps;
308 s->vlc_n = FFMIN(s->n, MAX_VLC_N);
309
310 avctx->bits_per_coded_sample = s->bitstream_bpp;
311 s->decorrelate = s->bitstream_bpp >= 24 && !s->yuv && !(desc->flags & AV_PIX_FMT_FLAG_PLANAR);
312 #if FF_API_PRIVATE_OPT
313 FF_DISABLE_DEPRECATION_WARNINGS
314 if (avctx->prediction_method)
315 s->predictor = avctx->prediction_method;
316 FF_ENABLE_DEPRECATION_WARNINGS
317 #endif
318 s->interlaced = avctx->flags & AV_CODEC_FLAG_INTERLACED_ME ? 1 : 0;
319 if (s->context) {
320 if (s->flags & (AV_CODEC_FLAG_PASS1 | AV_CODEC_FLAG_PASS2)) {
321 av_log(avctx, AV_LOG_ERROR,
322 "context=1 is not compatible with "
323 "2 pass huffyuv encoding\n");
324 return AVERROR(EINVAL);
325 }
326 }
327
328 if (avctx->codec->id == AV_CODEC_ID_HUFFYUV) {
329 if (avctx->pix_fmt == AV_PIX_FMT_YUV420P) {
330 av_log(avctx, AV_LOG_ERROR,
331 "Error: YV12 is not supported by huffyuv; use "
332 "vcodec=ffvhuff or format=422p\n");
333 return AVERROR(EINVAL);
334 }
335 #if FF_API_PRIVATE_OPT
336 if (s->context) {
337 av_log(avctx, AV_LOG_ERROR,
338 "Error: per-frame huffman tables are not supported "
339 "by huffyuv; use vcodec=ffvhuff\n");
340 return AVERROR(EINVAL);
341 }
342 if (s->version > 2) {
343 av_log(avctx, AV_LOG_ERROR,
344 "Error: ver>2 is not supported "
345 "by huffyuv; use vcodec=ffvhuff\n");
346 return AVERROR(EINVAL);
347 }
348 #endif
349 if (s->interlaced != ( s->height > 288 ))
350 av_log(avctx, AV_LOG_INFO,
351 "using huffyuv 2.2.0 or newer interlacing flag\n");
352 }
353
354 if (s->version > 3 && avctx->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL) {
355 av_log(avctx, AV_LOG_ERROR, "Ver > 3 is under development, files encoded with it may not be decodable with future versions!!!\n"
356 "Use vstrict=-2 / -strict -2 to use it anyway.\n");
357 return AVERROR(EINVAL);
358 }
359
360 if (s->bitstream_bpp >= 24 && s->predictor == MEDIAN && s->version <= 2) {
361 av_log(avctx, AV_LOG_ERROR,
362 "Error: RGB is incompatible with median predictor\n");
363 return AVERROR(EINVAL);
364 }
365
366 ((uint8_t*)avctx->extradata)[0] = s->predictor | (s->decorrelate << 6);
367 ((uint8_t*)avctx->extradata)[2] = s->interlaced ? 0x10 : 0x20;
368 if (s->context)
369 ((uint8_t*)avctx->extradata)[2] |= 0x40;
370 if (s->version < 3) {
371 ((uint8_t*)avctx->extradata)[1] = s->bitstream_bpp;
372 ((uint8_t*)avctx->extradata)[3] = 0;
373 } else {
374 ((uint8_t*)avctx->extradata)[1] = ((s->bps-1)<<4) | s->chroma_h_shift | (s->chroma_v_shift<<2);
375 if (s->chroma)
376 ((uint8_t*)avctx->extradata)[2] |= s->yuv ? 1 : 2;
377 if (s->alpha)
378 ((uint8_t*)avctx->extradata)[2] |= 4;
379 ((uint8_t*)avctx->extradata)[3] = 1;
380 }
381 s->avctx->extradata_size = 4;
382
383 if (avctx->stats_in) {
384 char *p = avctx->stats_in;
385
386 for (i = 0; i < 4; i++)
387 for (j = 0; j < s->vlc_n; j++)
388 s->stats[i][j] = 1;
389
390 for (;;) {
391 for (i = 0; i < 4; i++) {
392 char *next;
393
394 for (j = 0; j < s->vlc_n; j++) {
395 s->stats[i][j] += strtol(p, &next, 0);
396 if (next == p) return -1;
397 p = next;
398 }
399 }
400 if (p[0] == 0 || p[1] == 0 || p[2] == 0) break;
401 }
402 } else {
403 for (i = 0; i < 4; i++)
404 for (j = 0; j < s->vlc_n; j++) {
405 int d = FFMIN(j, s->vlc_n - j);
406
407 s->stats[i][j] = 100000000 / (d*d + 1);
408 }
409 }
410
411 ret = store_huffman_tables(s, s->avctx->extradata + s->avctx->extradata_size);
412 if (ret < 0)
413 return ret;
414 s->avctx->extradata_size += ret;
415
416 if (s->context) {
417 for (i = 0; i < 4; i++) {
418 int pels = s->width * s->height / (i ? 40 : 10);
419 for (j = 0; j < s->vlc_n; j++) {
420 int d = FFMIN(j, s->vlc_n - j);
421 s->stats[i][j] = pels/(d*d + 1);
422 }
423 }
424 } else {
425 for (i = 0; i < 4; i++)
426 for (j = 0; j < s->vlc_n; j++)
427 s->stats[i][j]= 0;
428 }
429
430 if (ff_huffyuv_alloc_temp(s)) {
431 ff_huffyuv_common_end(s);
432 return AVERROR(ENOMEM);
433 }
434
435 s->picture_number=0;
436
437 return 0;
438 }
encode_422_bitstream(HYuvContext * s,int offset,int count)439 static int encode_422_bitstream(HYuvContext *s, int offset, int count)
440 {
441 int i;
442 const uint8_t *y = s->temp[0] + offset;
443 const uint8_t *u = s->temp[1] + offset / 2;
444 const uint8_t *v = s->temp[2] + offset / 2;
445
446 if (s->pb.buf_end - s->pb.buf - (put_bits_count(&s->pb) >> 3) < 2 * 4 * count) {
447 av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
448 return -1;
449 }
450
451 #define LOAD4\
452 int y0 = y[2 * i];\
453 int y1 = y[2 * i + 1];\
454 int u0 = u[i];\
455 int v0 = v[i];
456
457 count /= 2;
458
459 if (s->flags & AV_CODEC_FLAG_PASS1) {
460 for(i = 0; i < count; i++) {
461 LOAD4;
462 s->stats[0][y0]++;
463 s->stats[1][u0]++;
464 s->stats[0][y1]++;
465 s->stats[2][v0]++;
466 }
467 }
468 if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
469 return 0;
470 if (s->context) {
471 for (i = 0; i < count; i++) {
472 LOAD4;
473 s->stats[0][y0]++;
474 put_bits(&s->pb, s->len[0][y0], s->bits[0][y0]);
475 s->stats[1][u0]++;
476 put_bits(&s->pb, s->len[1][u0], s->bits[1][u0]);
477 s->stats[0][y1]++;
478 put_bits(&s->pb, s->len[0][y1], s->bits[0][y1]);
479 s->stats[2][v0]++;
480 put_bits(&s->pb, s->len[2][v0], s->bits[2][v0]);
481 }
482 } else {
483 for(i = 0; i < count; i++) {
484 LOAD4;
485 put_bits(&s->pb, s->len[0][y0], s->bits[0][y0]);
486 put_bits(&s->pb, s->len[1][u0], s->bits[1][u0]);
487 put_bits(&s->pb, s->len[0][y1], s->bits[0][y1]);
488 put_bits(&s->pb, s->len[2][v0], s->bits[2][v0]);
489 }
490 }
491 return 0;
492 }
493
encode_plane_bitstream(HYuvContext * s,int width,int plane)494 static int encode_plane_bitstream(HYuvContext *s, int width, int plane)
495 {
496 int i, count = width/2;
497
498 if (s->pb.buf_end - s->pb.buf - (put_bits_count(&s->pb) >> 3) < count * s->bps / 2) {
499 av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
500 return -1;
501 }
502
503 #define LOADEND\
504 int y0 = s->temp[0][width-1];
505 #define LOADEND_14\
506 int y0 = s->temp16[0][width-1] & mask;
507 #define LOADEND_16\
508 int y0 = s->temp16[0][width-1];
509 #define STATEND\
510 s->stats[plane][y0]++;
511 #define STATEND_16\
512 s->stats[plane][y0>>2]++;
513 #define WRITEEND\
514 put_bits(&s->pb, s->len[plane][y0], s->bits[plane][y0]);
515 #define WRITEEND_16\
516 put_bits(&s->pb, s->len[plane][y0>>2], s->bits[plane][y0>>2]);\
517 put_bits(&s->pb, 2, y0&3);
518
519 #define LOAD2\
520 int y0 = s->temp[0][2 * i];\
521 int y1 = s->temp[0][2 * i + 1];
522 #define LOAD2_14\
523 int y0 = s->temp16[0][2 * i] & mask;\
524 int y1 = s->temp16[0][2 * i + 1] & mask;
525 #define LOAD2_16\
526 int y0 = s->temp16[0][2 * i];\
527 int y1 = s->temp16[0][2 * i + 1];
528 #define STAT2\
529 s->stats[plane][y0]++;\
530 s->stats[plane][y1]++;
531 #define STAT2_16\
532 s->stats[plane][y0>>2]++;\
533 s->stats[plane][y1>>2]++;
534 #define WRITE2\
535 put_bits(&s->pb, s->len[plane][y0], s->bits[plane][y0]);\
536 put_bits(&s->pb, s->len[plane][y1], s->bits[plane][y1]);
537 #define WRITE2_16\
538 put_bits(&s->pb, s->len[plane][y0>>2], s->bits[plane][y0>>2]);\
539 put_bits(&s->pb, 2, y0&3);\
540 put_bits(&s->pb, s->len[plane][y1>>2], s->bits[plane][y1>>2]);\
541 put_bits(&s->pb, 2, y1&3);
542
543 if (s->bps <= 8) {
544 if (s->flags & AV_CODEC_FLAG_PASS1) {
545 for (i = 0; i < count; i++) {
546 LOAD2;
547 STAT2;
548 }
549 if (width&1) {
550 LOADEND;
551 STATEND;
552 }
553 }
554 if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
555 return 0;
556
557 if (s->context) {
558 for (i = 0; i < count; i++) {
559 LOAD2;
560 STAT2;
561 WRITE2;
562 }
563 if (width&1) {
564 LOADEND;
565 STATEND;
566 WRITEEND;
567 }
568 } else {
569 for (i = 0; i < count; i++) {
570 LOAD2;
571 WRITE2;
572 }
573 if (width&1) {
574 LOADEND;
575 WRITEEND;
576 }
577 }
578 } else if (s->bps <= 14) {
579 int mask = s->n - 1;
580 if (s->flags & AV_CODEC_FLAG_PASS1) {
581 for (i = 0; i < count; i++) {
582 LOAD2_14;
583 STAT2;
584 }
585 if (width&1) {
586 LOADEND_14;
587 STATEND;
588 }
589 }
590 if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
591 return 0;
592
593 if (s->context) {
594 for (i = 0; i < count; i++) {
595 LOAD2_14;
596 STAT2;
597 WRITE2;
598 }
599 if (width&1) {
600 LOADEND_14;
601 STATEND;
602 WRITEEND;
603 }
604 } else {
605 for (i = 0; i < count; i++) {
606 LOAD2_14;
607 WRITE2;
608 }
609 if (width&1) {
610 LOADEND_14;
611 WRITEEND;
612 }
613 }
614 } else {
615 if (s->flags & AV_CODEC_FLAG_PASS1) {
616 for (i = 0; i < count; i++) {
617 LOAD2_16;
618 STAT2_16;
619 }
620 if (width&1) {
621 LOADEND_16;
622 STATEND_16;
623 }
624 }
625 if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
626 return 0;
627
628 if (s->context) {
629 for (i = 0; i < count; i++) {
630 LOAD2_16;
631 STAT2_16;
632 WRITE2_16;
633 }
634 if (width&1) {
635 LOADEND_16;
636 STATEND_16;
637 WRITEEND_16;
638 }
639 } else {
640 for (i = 0; i < count; i++) {
641 LOAD2_16;
642 WRITE2_16;
643 }
644 if (width&1) {
645 LOADEND_16;
646 WRITEEND_16;
647 }
648 }
649 }
650 #undef LOAD2
651 #undef STAT2
652 #undef WRITE2
653 return 0;
654 }
655
encode_gray_bitstream(HYuvContext * s,int count)656 static int encode_gray_bitstream(HYuvContext *s, int count)
657 {
658 int i;
659
660 if (s->pb.buf_end - s->pb.buf - (put_bits_count(&s->pb) >> 3) < 4 * count) {
661 av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
662 return -1;
663 }
664
665 #define LOAD2\
666 int y0 = s->temp[0][2 * i];\
667 int y1 = s->temp[0][2 * i + 1];
668 #define STAT2\
669 s->stats[0][y0]++;\
670 s->stats[0][y1]++;
671 #define WRITE2\
672 put_bits(&s->pb, s->len[0][y0], s->bits[0][y0]);\
673 put_bits(&s->pb, s->len[0][y1], s->bits[0][y1]);
674
675 count /= 2;
676
677 if (s->flags & AV_CODEC_FLAG_PASS1) {
678 for (i = 0; i < count; i++) {
679 LOAD2;
680 STAT2;
681 }
682 }
683 if (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)
684 return 0;
685
686 if (s->context) {
687 for (i = 0; i < count; i++) {
688 LOAD2;
689 STAT2;
690 WRITE2;
691 }
692 } else {
693 for (i = 0; i < count; i++) {
694 LOAD2;
695 WRITE2;
696 }
697 }
698 return 0;
699 }
700
encode_bgra_bitstream(HYuvContext * s,int count,int planes)701 static inline int encode_bgra_bitstream(HYuvContext *s, int count, int planes)
702 {
703 int i;
704
705 if (s->pb.buf_end - s->pb.buf - (put_bits_count(&s->pb) >> 3) <
706 4 * planes * count) {
707 av_log(s->avctx, AV_LOG_ERROR, "encoded frame too large\n");
708 return -1;
709 }
710
711 #define LOAD_GBRA \
712 int g = s->temp[0][planes == 3 ? 3 * i + 1 : 4 * i + G]; \
713 int b =(s->temp[0][planes == 3 ? 3 * i + 2 : 4 * i + B] - g) & 0xFF;\
714 int r =(s->temp[0][planes == 3 ? 3 * i + 0 : 4 * i + R] - g) & 0xFF;\
715 int a = s->temp[0][planes * i + A];
716
717 #define STAT_BGRA \
718 s->stats[0][b]++; \
719 s->stats[1][g]++; \
720 s->stats[2][r]++; \
721 if (planes == 4) \
722 s->stats[2][a]++;
723
724 #define WRITE_GBRA \
725 put_bits(&s->pb, s->len[1][g], s->bits[1][g]); \
726 put_bits(&s->pb, s->len[0][b], s->bits[0][b]); \
727 put_bits(&s->pb, s->len[2][r], s->bits[2][r]); \
728 if (planes == 4) \
729 put_bits(&s->pb, s->len[2][a], s->bits[2][a]);
730
731 if ((s->flags & AV_CODEC_FLAG_PASS1) &&
732 (s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)) {
733 for (i = 0; i < count; i++) {
734 LOAD_GBRA;
735 STAT_BGRA;
736 }
737 } else if (s->context || (s->flags & AV_CODEC_FLAG_PASS1)) {
738 for (i = 0; i < count; i++) {
739 LOAD_GBRA;
740 STAT_BGRA;
741 WRITE_GBRA;
742 }
743 } else {
744 for (i = 0; i < count; i++) {
745 LOAD_GBRA;
746 WRITE_GBRA;
747 }
748 }
749 return 0;
750 }
751
encode_frame(AVCodecContext * avctx,AVPacket * pkt,const AVFrame * pict,int * got_packet)752 static int encode_frame(AVCodecContext *avctx, AVPacket *pkt,
753 const AVFrame *pict, int *got_packet)
754 {
755 HYuvContext *s = avctx->priv_data;
756 const int width = s->width;
757 const int width2 = s->width>>1;
758 const int height = s->height;
759 const int fake_ystride = s->interlaced ? pict->linesize[0]*2 : pict->linesize[0];
760 const int fake_ustride = s->interlaced ? pict->linesize[1]*2 : pict->linesize[1];
761 const int fake_vstride = s->interlaced ? pict->linesize[2]*2 : pict->linesize[2];
762 const AVFrame * const p = pict;
763 int i, j, size = 0, ret;
764
765 if ((ret = ff_alloc_packet2(avctx, pkt, width * height * 3 * 4 + AV_INPUT_BUFFER_MIN_SIZE, 0)) < 0)
766 return ret;
767
768 if (s->context) {
769 size = store_huffman_tables(s, pkt->data);
770 if (size < 0)
771 return size;
772
773 for (i = 0; i < 4; i++)
774 for (j = 0; j < s->vlc_n; j++)
775 s->stats[i][j] >>= 1;
776 }
777
778 init_put_bits(&s->pb, pkt->data + size, pkt->size - size);
779
780 if (avctx->pix_fmt == AV_PIX_FMT_YUV422P ||
781 avctx->pix_fmt == AV_PIX_FMT_YUV420P) {
782 int lefty, leftu, leftv, y, cy;
783
784 put_bits(&s->pb, 8, leftv = p->data[2][0]);
785 put_bits(&s->pb, 8, lefty = p->data[0][1]);
786 put_bits(&s->pb, 8, leftu = p->data[1][0]);
787 put_bits(&s->pb, 8, p->data[0][0]);
788
789 lefty = sub_left_prediction(s, s->temp[0], p->data[0], width , 0);
790 leftu = sub_left_prediction(s, s->temp[1], p->data[1], width2, 0);
791 leftv = sub_left_prediction(s, s->temp[2], p->data[2], width2, 0);
792
793 encode_422_bitstream(s, 2, width-2);
794
795 if (s->predictor==MEDIAN) {
796 int lefttopy, lefttopu, lefttopv;
797 cy = y = 1;
798 if (s->interlaced) {
799 lefty = sub_left_prediction(s, s->temp[0], p->data[0] + p->linesize[0], width , lefty);
800 leftu = sub_left_prediction(s, s->temp[1], p->data[1] + p->linesize[1], width2, leftu);
801 leftv = sub_left_prediction(s, s->temp[2], p->data[2] + p->linesize[2], width2, leftv);
802
803 encode_422_bitstream(s, 0, width);
804 y++; cy++;
805 }
806
807 lefty = sub_left_prediction(s, s->temp[0], p->data[0] + fake_ystride, 4, lefty);
808 leftu = sub_left_prediction(s, s->temp[1], p->data[1] + fake_ustride, 2, leftu);
809 leftv = sub_left_prediction(s, s->temp[2], p->data[2] + fake_vstride, 2, leftv);
810
811 encode_422_bitstream(s, 0, 4);
812
813 lefttopy = p->data[0][3];
814 lefttopu = p->data[1][1];
815 lefttopv = p->data[2][1];
816 s->llvidencdsp.sub_median_pred(s->temp[0], p->data[0] + 4, p->data[0] + fake_ystride + 4, width - 4, &lefty, &lefttopy);
817 s->llvidencdsp.sub_median_pred(s->temp[1], p->data[1] + 2, p->data[1] + fake_ustride + 2, width2 - 2, &leftu, &lefttopu);
818 s->llvidencdsp.sub_median_pred(s->temp[2], p->data[2] + 2, p->data[2] + fake_vstride + 2, width2 - 2, &leftv, &lefttopv);
819 encode_422_bitstream(s, 0, width - 4);
820 y++; cy++;
821
822 for (; y < height; y++,cy++) {
823 uint8_t *ydst, *udst, *vdst;
824
825 if (s->bitstream_bpp == 12) {
826 while (2 * cy > y) {
827 ydst = p->data[0] + p->linesize[0] * y;
828 s->llvidencdsp.sub_median_pred(s->temp[0], ydst - fake_ystride, ydst, width, &lefty, &lefttopy);
829 encode_gray_bitstream(s, width);
830 y++;
831 }
832 if (y >= height) break;
833 }
834 ydst = p->data[0] + p->linesize[0] * y;
835 udst = p->data[1] + p->linesize[1] * cy;
836 vdst = p->data[2] + p->linesize[2] * cy;
837
838 s->llvidencdsp.sub_median_pred(s->temp[0], ydst - fake_ystride, ydst, width, &lefty, &lefttopy);
839 s->llvidencdsp.sub_median_pred(s->temp[1], udst - fake_ustride, udst, width2, &leftu, &lefttopu);
840 s->llvidencdsp.sub_median_pred(s->temp[2], vdst - fake_vstride, vdst, width2, &leftv, &lefttopv);
841
842 encode_422_bitstream(s, 0, width);
843 }
844 } else {
845 for (cy = y = 1; y < height; y++, cy++) {
846 uint8_t *ydst, *udst, *vdst;
847
848 /* encode a luma only line & y++ */
849 if (s->bitstream_bpp == 12) {
850 ydst = p->data[0] + p->linesize[0] * y;
851
852 if (s->predictor == PLANE && s->interlaced < y) {
853 s->llvidencdsp.diff_bytes(s->temp[1], ydst, ydst - fake_ystride, width);
854
855 lefty = sub_left_prediction(s, s->temp[0], s->temp[1], width , lefty);
856 } else {
857 lefty = sub_left_prediction(s, s->temp[0], ydst, width , lefty);
858 }
859 encode_gray_bitstream(s, width);
860 y++;
861 if (y >= height) break;
862 }
863
864 ydst = p->data[0] + p->linesize[0] * y;
865 udst = p->data[1] + p->linesize[1] * cy;
866 vdst = p->data[2] + p->linesize[2] * cy;
867
868 if (s->predictor == PLANE && s->interlaced < cy) {
869 s->llvidencdsp.diff_bytes(s->temp[1], ydst, ydst - fake_ystride, width);
870 s->llvidencdsp.diff_bytes(s->temp[2], udst, udst - fake_ustride, width2);
871 s->llvidencdsp.diff_bytes(s->temp[2] + width2, vdst, vdst - fake_vstride, width2);
872
873 lefty = sub_left_prediction(s, s->temp[0], s->temp[1], width , lefty);
874 leftu = sub_left_prediction(s, s->temp[1], s->temp[2], width2, leftu);
875 leftv = sub_left_prediction(s, s->temp[2], s->temp[2] + width2, width2, leftv);
876 } else {
877 lefty = sub_left_prediction(s, s->temp[0], ydst, width , lefty);
878 leftu = sub_left_prediction(s, s->temp[1], udst, width2, leftu);
879 leftv = sub_left_prediction(s, s->temp[2], vdst, width2, leftv);
880 }
881
882 encode_422_bitstream(s, 0, width);
883 }
884 }
885 } else if(avctx->pix_fmt == AV_PIX_FMT_RGB32) {
886 uint8_t *data = p->data[0] + (height - 1) * p->linesize[0];
887 const int stride = -p->linesize[0];
888 const int fake_stride = -fake_ystride;
889 int y;
890 int leftr, leftg, leftb, lefta;
891
892 put_bits(&s->pb, 8, lefta = data[A]);
893 put_bits(&s->pb, 8, leftr = data[R]);
894 put_bits(&s->pb, 8, leftg = data[G]);
895 put_bits(&s->pb, 8, leftb = data[B]);
896
897 sub_left_prediction_bgr32(s, s->temp[0], data + 4, width - 1,
898 &leftr, &leftg, &leftb, &lefta);
899 encode_bgra_bitstream(s, width - 1, 4);
900
901 for (y = 1; y < s->height; y++) {
902 uint8_t *dst = data + y*stride;
903 if (s->predictor == PLANE && s->interlaced < y) {
904 s->llvidencdsp.diff_bytes(s->temp[1], dst, dst - fake_stride, width * 4);
905 sub_left_prediction_bgr32(s, s->temp[0], s->temp[1], width,
906 &leftr, &leftg, &leftb, &lefta);
907 } else {
908 sub_left_prediction_bgr32(s, s->temp[0], dst, width,
909 &leftr, &leftg, &leftb, &lefta);
910 }
911 encode_bgra_bitstream(s, width, 4);
912 }
913 } else if (avctx->pix_fmt == AV_PIX_FMT_RGB24) {
914 uint8_t *data = p->data[0] + (height - 1) * p->linesize[0];
915 const int stride = -p->linesize[0];
916 const int fake_stride = -fake_ystride;
917 int y;
918 int leftr, leftg, leftb;
919
920 put_bits(&s->pb, 8, leftr = data[0]);
921 put_bits(&s->pb, 8, leftg = data[1]);
922 put_bits(&s->pb, 8, leftb = data[2]);
923 put_bits(&s->pb, 8, 0);
924
925 sub_left_prediction_rgb24(s, s->temp[0], data + 3, width - 1,
926 &leftr, &leftg, &leftb);
927 encode_bgra_bitstream(s, width-1, 3);
928
929 for (y = 1; y < s->height; y++) {
930 uint8_t *dst = data + y * stride;
931 if (s->predictor == PLANE && s->interlaced < y) {
932 s->llvidencdsp.diff_bytes(s->temp[1], dst, dst - fake_stride,
933 width * 3);
934 sub_left_prediction_rgb24(s, s->temp[0], s->temp[1], width,
935 &leftr, &leftg, &leftb);
936 } else {
937 sub_left_prediction_rgb24(s, s->temp[0], dst, width,
938 &leftr, &leftg, &leftb);
939 }
940 encode_bgra_bitstream(s, width, 3);
941 }
942 } else if (s->version > 2) {
943 int plane;
944 for (plane = 0; plane < 1 + 2*s->chroma + s->alpha; plane++) {
945 int left, y;
946 int w = width;
947 int h = height;
948 int fake_stride = fake_ystride;
949
950 if (s->chroma && (plane == 1 || plane == 2)) {
951 w >>= s->chroma_h_shift;
952 h >>= s->chroma_v_shift;
953 fake_stride = plane == 1 ? fake_ustride : fake_vstride;
954 }
955
956 left = sub_left_prediction(s, s->temp[0], p->data[plane], w , 0);
957
958 encode_plane_bitstream(s, w, plane);
959
960 if (s->predictor==MEDIAN) {
961 int lefttop;
962 y = 1;
963 if (s->interlaced) {
964 left = sub_left_prediction(s, s->temp[0], p->data[plane] + p->linesize[plane], w , left);
965
966 encode_plane_bitstream(s, w, plane);
967 y++;
968 }
969
970 lefttop = p->data[plane][0];
971
972 for (; y < h; y++) {
973 uint8_t *dst = p->data[plane] + p->linesize[plane] * y;
974
975 sub_median_prediction(s, s->temp[0], dst - fake_stride, dst, w , &left, &lefttop);
976
977 encode_plane_bitstream(s, w, plane);
978 }
979 } else {
980 for (y = 1; y < h; y++) {
981 uint8_t *dst = p->data[plane] + p->linesize[plane] * y;
982
983 if (s->predictor == PLANE && s->interlaced < y) {
984 diff_bytes(s, s->temp[1], dst, dst - fake_stride, w);
985
986 left = sub_left_prediction(s, s->temp[0], s->temp[1], w , left);
987 } else {
988 left = sub_left_prediction(s, s->temp[0], dst, w , left);
989 }
990
991 encode_plane_bitstream(s, w, plane);
992 }
993 }
994 }
995 } else {
996 av_log(avctx, AV_LOG_ERROR, "Format not supported!\n");
997 }
998 emms_c();
999
1000 size += (put_bits_count(&s->pb) + 31) / 8;
1001 put_bits(&s->pb, 16, 0);
1002 put_bits(&s->pb, 15, 0);
1003 size /= 4;
1004
1005 if ((s->flags & AV_CODEC_FLAG_PASS1) && (s->picture_number & 31) == 0) {
1006 int j;
1007 char *p = avctx->stats_out;
1008 char *end = p + STATS_OUT_SIZE;
1009 for (i = 0; i < 4; i++) {
1010 for (j = 0; j < s->vlc_n; j++) {
1011 snprintf(p, end-p, "%"PRIu64" ", s->stats[i][j]);
1012 p += strlen(p);
1013 s->stats[i][j]= 0;
1014 }
1015 snprintf(p, end-p, "\n");
1016 p++;
1017 if (end <= p)
1018 return AVERROR(ENOMEM);
1019 }
1020 } else if (avctx->stats_out)
1021 avctx->stats_out[0] = '\0';
1022 if (!(s->avctx->flags2 & AV_CODEC_FLAG2_NO_OUTPUT)) {
1023 flush_put_bits(&s->pb);
1024 s->bdsp.bswap_buf((uint32_t *) pkt->data, (uint32_t *) pkt->data, size);
1025 }
1026
1027 s->picture_number++;
1028
1029 pkt->size = size * 4;
1030 pkt->flags |= AV_PKT_FLAG_KEY;
1031 *got_packet = 1;
1032
1033 return 0;
1034 }
1035
encode_end(AVCodecContext * avctx)1036 static av_cold int encode_end(AVCodecContext *avctx)
1037 {
1038 HYuvContext *s = avctx->priv_data;
1039
1040 ff_huffyuv_common_end(s);
1041
1042 av_freep(&avctx->extradata);
1043 av_freep(&avctx->stats_out);
1044
1045 return 0;
1046 }
1047
1048 #define OFFSET(x) offsetof(HYuvContext, x)
1049 #define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
1050
1051 #define COMMON_OPTIONS \
1052 { "non_deterministic", "Allow multithreading for e.g. context=1 at the expense of determinism", \
1053 OFFSET(non_determ), AV_OPT_TYPE_BOOL, { .i64 = 1 }, \
1054 0, 1, VE }, \
1055 { "pred", "Prediction method", OFFSET(predictor), AV_OPT_TYPE_INT, { .i64 = LEFT }, LEFT, MEDIAN, VE, "pred" }, \
1056 { "left", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = LEFT }, INT_MIN, INT_MAX, VE, "pred" }, \
1057 { "plane", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = PLANE }, INT_MIN, INT_MAX, VE, "pred" }, \
1058 { "median", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = MEDIAN }, INT_MIN, INT_MAX, VE, "pred" }, \
1059
1060 static const AVOption normal_options[] = {
1061 COMMON_OPTIONS
1062 { NULL },
1063 };
1064
1065 static const AVOption ff_options[] = {
1066 COMMON_OPTIONS
1067 { "context", "Set per-frame huffman tables", OFFSET(context), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE },
1068 { NULL },
1069 };
1070
1071 static const AVClass normal_class = {
1072 .class_name = "huffyuv",
1073 .item_name = av_default_item_name,
1074 .option = normal_options,
1075 .version = LIBAVUTIL_VERSION_INT,
1076 };
1077
1078 static const AVClass ff_class = {
1079 .class_name = "ffvhuff",
1080 .item_name = av_default_item_name,
1081 .option = ff_options,
1082 .version = LIBAVUTIL_VERSION_INT,
1083 };
1084
1085 AVCodec ff_huffyuv_encoder = {
1086 .name = "huffyuv",
1087 .long_name = NULL_IF_CONFIG_SMALL("Huffyuv / HuffYUV"),
1088 .type = AVMEDIA_TYPE_VIDEO,
1089 .id = AV_CODEC_ID_HUFFYUV,
1090 .priv_data_size = sizeof(HYuvContext),
1091 .init = encode_init,
1092 .encode2 = encode_frame,
1093 .close = encode_end,
1094 .capabilities = AV_CODEC_CAP_FRAME_THREADS,
1095 .priv_class = &normal_class,
1096 .pix_fmts = (const enum AVPixelFormat[]){
1097 AV_PIX_FMT_YUV422P, AV_PIX_FMT_RGB24,
1098 AV_PIX_FMT_RGB32, AV_PIX_FMT_NONE
1099 },
1100 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE |
1101 FF_CODEC_CAP_INIT_CLEANUP,
1102 };
1103
1104 #if CONFIG_FFVHUFF_ENCODER
1105 AVCodec ff_ffvhuff_encoder = {
1106 .name = "ffvhuff",
1107 .long_name = NULL_IF_CONFIG_SMALL("Huffyuv FFmpeg variant"),
1108 .type = AVMEDIA_TYPE_VIDEO,
1109 .id = AV_CODEC_ID_FFVHUFF,
1110 .priv_data_size = sizeof(HYuvContext),
1111 .init = encode_init,
1112 .encode2 = encode_frame,
1113 .close = encode_end,
1114 .capabilities = AV_CODEC_CAP_FRAME_THREADS,
1115 .priv_class = &ff_class,
1116 .pix_fmts = (const enum AVPixelFormat[]){
1117 AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV411P,
1118 AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUV440P,
1119 AV_PIX_FMT_GBRP,
1120 AV_PIX_FMT_GBRP9, AV_PIX_FMT_GBRP10, AV_PIX_FMT_GBRP12, AV_PIX_FMT_GBRP14, AV_PIX_FMT_GBRP16,
1121 AV_PIX_FMT_GRAY8, AV_PIX_FMT_GRAY16,
1122 AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUVA444P,
1123 AV_PIX_FMT_GBRAP,
1124 AV_PIX_FMT_YUV420P9, AV_PIX_FMT_YUV420P10, AV_PIX_FMT_YUV420P12, AV_PIX_FMT_YUV420P14, AV_PIX_FMT_YUV420P16,
1125 AV_PIX_FMT_YUV422P9, AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUV422P12, AV_PIX_FMT_YUV422P14, AV_PIX_FMT_YUV422P16,
1126 AV_PIX_FMT_YUV444P9, AV_PIX_FMT_YUV444P10, AV_PIX_FMT_YUV444P12, AV_PIX_FMT_YUV444P14, AV_PIX_FMT_YUV444P16,
1127 AV_PIX_FMT_YUVA420P9, AV_PIX_FMT_YUVA420P10, AV_PIX_FMT_YUVA420P16,
1128 AV_PIX_FMT_YUVA422P9, AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_YUVA422P16,
1129 AV_PIX_FMT_YUVA444P9, AV_PIX_FMT_YUVA444P10, AV_PIX_FMT_YUVA444P16,
1130 AV_PIX_FMT_RGB24,
1131 AV_PIX_FMT_RGB32, AV_PIX_FMT_NONE
1132 },
1133 .caps_internal = FF_CODEC_CAP_INIT_THREADSAFE |
1134 FF_CODEC_CAP_INIT_CLEANUP,
1135 };
1136 #endif
1137