Commit | Line | Data |
---|---|---|
2ba45a60 DM |
1 | /* |
2 | * OpenEXR (.exr) image decoder | |
3 | * Copyright (c) 2009 Jimmy Christensen | |
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 | * OpenEXR decoder | |
25 | * @author Jimmy Christensen | |
26 | * | |
27 | * For more information on the OpenEXR format, visit: | |
28 | * http://openexr.com/ | |
29 | * | |
30 | * exr_flt2uint() and exr_halflt2uint() is credited to Reimar Döffinger. | |
31 | * exr_half2float() is credited to Aaftab Munshi; Dan Ginsburg, Dave Shreiner. | |
32 | * | |
33 | */ | |
34 | ||
35 | #include <zlib.h> | |
36 | #include <float.h> | |
37 | ||
38 | #include "libavutil/imgutils.h" | |
39 | #include "libavutil/opt.h" | |
40 | #include "libavutil/intfloat.h" | |
41 | ||
42 | #include "avcodec.h" | |
43 | #include "bytestream.h" | |
44 | #include "get_bits.h" | |
45 | #include "internal.h" | |
46 | #include "mathops.h" | |
47 | #include "thread.h" | |
48 | ||
49 | enum ExrCompr { | |
50 | EXR_RAW, | |
51 | EXR_RLE, | |
52 | EXR_ZIP1, | |
53 | EXR_ZIP16, | |
54 | EXR_PIZ, | |
55 | EXR_PXR24, | |
56 | EXR_B44, | |
57 | EXR_B44A, | |
58 | EXR_UNKN, | |
59 | }; | |
60 | ||
61 | enum ExrPixelType { | |
62 | EXR_UINT, | |
63 | EXR_HALF, | |
64 | EXR_FLOAT, | |
65 | EXR_UNKNOWN, | |
66 | }; | |
67 | ||
68 | typedef struct EXRChannel { | |
69 | int xsub, ysub; | |
70 | enum ExrPixelType pixel_type; | |
71 | } EXRChannel; | |
72 | ||
73 | typedef struct EXRThreadData { | |
74 | uint8_t *uncompressed_data; | |
75 | int uncompressed_size; | |
76 | ||
77 | uint8_t *tmp; | |
78 | int tmp_size; | |
79 | ||
80 | uint8_t *bitmap; | |
81 | uint16_t *lut; | |
82 | } EXRThreadData; | |
83 | ||
84 | typedef struct EXRContext { | |
85 | AVClass *class; | |
86 | AVFrame *picture; | |
87 | AVCodecContext *avctx; | |
88 | ||
89 | enum ExrCompr compression; | |
90 | enum ExrPixelType pixel_type; | |
91 | int channel_offsets[4]; // 0 = red, 1 = green, 2 = blue and 3 = alpha | |
92 | const AVPixFmtDescriptor *desc; | |
93 | ||
94 | int w, h; | |
95 | uint32_t xmax, xmin; | |
96 | uint32_t ymax, ymin; | |
97 | uint32_t xdelta, ydelta; | |
98 | int ysize; | |
99 | ||
100 | uint64_t scan_line_size; | |
101 | int scan_lines_per_block; | |
102 | ||
103 | GetByteContext gb; | |
104 | const uint8_t *buf; | |
105 | int buf_size; | |
106 | ||
107 | EXRChannel *channels; | |
108 | int nb_channels; | |
109 | ||
110 | EXRThreadData *thread_data; | |
111 | ||
112 | const char *layer; | |
113 | ||
114 | float gamma; | |
115 | ||
116 | uint16_t gamma_table[65536]; | |
117 | ||
118 | } EXRContext; | |
119 | ||
120 | /* -15 stored using a single precision bias of 127 */ | |
121 | #define HALF_FLOAT_MIN_BIASED_EXP_AS_SINGLE_FP_EXP 0x38000000 | |
122 | /* max exponent value in single precision that will be converted | |
123 | * to Inf or Nan when stored as a half-float */ | |
124 | #define HALF_FLOAT_MAX_BIASED_EXP_AS_SINGLE_FP_EXP 0x47800000 | |
125 | ||
126 | /* 255 is the max exponent biased value */ | |
127 | #define FLOAT_MAX_BIASED_EXP (0xFF << 23) | |
128 | ||
129 | #define HALF_FLOAT_MAX_BIASED_EXP (0x1F << 10) | |
130 | ||
131 | /* | |
132 | * Convert a half float as a uint16_t into a full float. | |
133 | * | |
134 | * @param hf half float as uint16_t | |
135 | * | |
136 | * @return float value | |
137 | */ | |
138 | static union av_intfloat32 exr_half2float(uint16_t hf) | |
139 | { | |
140 | unsigned int sign = (unsigned int)(hf >> 15); | |
141 | unsigned int mantissa = (unsigned int)(hf & ((1 << 10) - 1)); | |
142 | unsigned int exp = (unsigned int)(hf & HALF_FLOAT_MAX_BIASED_EXP); | |
143 | union av_intfloat32 f; | |
144 | ||
145 | if (exp == HALF_FLOAT_MAX_BIASED_EXP) { | |
146 | // we have a half-float NaN or Inf | |
147 | // half-float NaNs will be converted to a single precision NaN | |
148 | // half-float Infs will be converted to a single precision Inf | |
149 | exp = FLOAT_MAX_BIASED_EXP; | |
150 | if (mantissa) | |
151 | mantissa = (1 << 23) - 1; // set all bits to indicate a NaN | |
152 | } else if (exp == 0x0) { | |
153 | // convert half-float zero/denorm to single precision value | |
154 | if (mantissa) { | |
155 | mantissa <<= 1; | |
156 | exp = HALF_FLOAT_MIN_BIASED_EXP_AS_SINGLE_FP_EXP; | |
157 | // check for leading 1 in denorm mantissa | |
158 | while ((mantissa & (1 << 10))) { | |
159 | // for every leading 0, decrement single precision exponent by 1 | |
160 | // and shift half-float mantissa value to the left | |
161 | mantissa <<= 1; | |
162 | exp -= (1 << 23); | |
163 | } | |
164 | // clamp the mantissa to 10-bits | |
165 | mantissa &= ((1 << 10) - 1); | |
166 | // shift left to generate single-precision mantissa of 23-bits | |
167 | mantissa <<= 13; | |
168 | } | |
169 | } else { | |
170 | // shift left to generate single-precision mantissa of 23-bits | |
171 | mantissa <<= 13; | |
172 | // generate single precision biased exponent value | |
173 | exp = (exp << 13) + HALF_FLOAT_MIN_BIASED_EXP_AS_SINGLE_FP_EXP; | |
174 | } | |
175 | ||
176 | f.i = (sign << 31) | exp | mantissa; | |
177 | ||
178 | return f; | |
179 | } | |
180 | ||
181 | ||
182 | /** | |
183 | * Convert from 32-bit float as uint32_t to uint16_t. | |
184 | * | |
185 | * @param v 32-bit float | |
186 | * | |
187 | * @return normalized 16-bit unsigned int | |
188 | */ | |
189 | static inline uint16_t exr_flt2uint(uint32_t v) | |
190 | { | |
191 | unsigned int exp = v >> 23; | |
192 | // "HACK": negative values result in exp< 0, so clipping them to 0 | |
193 | // is also handled by this condition, avoids explicit check for sign bit. | |
194 | if (exp <= 127 + 7 - 24) // we would shift out all bits anyway | |
195 | return 0; | |
196 | if (exp >= 127) | |
197 | return 0xffff; | |
198 | v &= 0x007fffff; | |
199 | return (v + (1 << 23)) >> (127 + 7 - exp); | |
200 | } | |
201 | ||
202 | /** | |
203 | * Convert from 16-bit float as uint16_t to uint16_t. | |
204 | * | |
205 | * @param v 16-bit float | |
206 | * | |
207 | * @return normalized 16-bit unsigned int | |
208 | */ | |
209 | static inline uint16_t exr_halflt2uint(uint16_t v) | |
210 | { | |
211 | unsigned exp = 14 - (v >> 10); | |
212 | if (exp >= 14) { | |
213 | if (exp == 14) | |
214 | return (v >> 9) & 1; | |
215 | else | |
216 | return (v & 0x8000) ? 0 : 0xffff; | |
217 | } | |
218 | v <<= 6; | |
219 | return (v + (1 << 16)) >> (exp + 1); | |
220 | } | |
221 | ||
222 | static void predictor(uint8_t *src, int size) | |
223 | { | |
224 | uint8_t *t = src + 1; | |
225 | uint8_t *stop = src + size; | |
226 | ||
227 | while (t < stop) { | |
228 | int d = (int) t[-1] + (int) t[0] - 128; | |
229 | t[0] = d; | |
230 | ++t; | |
231 | } | |
232 | } | |
233 | ||
234 | static void reorder_pixels(uint8_t *src, uint8_t *dst, int size) | |
235 | { | |
236 | const int8_t *t1 = src; | |
237 | const int8_t *t2 = src + (size + 1) / 2; | |
238 | int8_t *s = dst; | |
239 | int8_t *stop = s + size; | |
240 | ||
241 | while (1) { | |
242 | if (s < stop) | |
243 | *(s++) = *(t1++); | |
244 | else | |
245 | break; | |
246 | ||
247 | if (s < stop) | |
248 | *(s++) = *(t2++); | |
249 | else | |
250 | break; | |
251 | } | |
252 | } | |
253 | ||
254 | static int zip_uncompress(const uint8_t *src, int compressed_size, | |
255 | int uncompressed_size, EXRThreadData *td) | |
256 | { | |
257 | unsigned long dest_len = uncompressed_size; | |
258 | ||
259 | if (uncompress(td->tmp, &dest_len, src, compressed_size) != Z_OK || | |
260 | dest_len != uncompressed_size) | |
261 | return AVERROR_INVALIDDATA; | |
262 | ||
263 | predictor(td->tmp, uncompressed_size); | |
264 | reorder_pixels(td->tmp, td->uncompressed_data, uncompressed_size); | |
265 | ||
266 | return 0; | |
267 | } | |
268 | ||
269 | static int rle_uncompress(const uint8_t *src, int compressed_size, | |
270 | int uncompressed_size, EXRThreadData *td) | |
271 | { | |
272 | uint8_t *d = td->tmp; | |
273 | const int8_t *s = src; | |
274 | int ssize = compressed_size; | |
275 | int dsize = uncompressed_size; | |
276 | uint8_t *dend = d + dsize; | |
277 | int count; | |
278 | ||
279 | while (ssize > 0) { | |
280 | count = *s++; | |
281 | ||
282 | if (count < 0) { | |
283 | count = -count; | |
284 | ||
285 | if ((dsize -= count) < 0 || | |
286 | (ssize -= count + 1) < 0) | |
287 | return AVERROR_INVALIDDATA; | |
288 | ||
289 | while (count--) | |
290 | *d++ = *s++; | |
291 | } else { | |
292 | count++; | |
293 | ||
294 | if ((dsize -= count) < 0 || | |
295 | (ssize -= 2) < 0) | |
296 | return AVERROR_INVALIDDATA; | |
297 | ||
298 | while (count--) | |
299 | *d++ = *s; | |
300 | ||
301 | s++; | |
302 | } | |
303 | } | |
304 | ||
305 | if (dend != d) | |
306 | return AVERROR_INVALIDDATA; | |
307 | ||
308 | predictor(td->tmp, uncompressed_size); | |
309 | reorder_pixels(td->tmp, td->uncompressed_data, uncompressed_size); | |
310 | ||
311 | return 0; | |
312 | } | |
313 | ||
314 | #define USHORT_RANGE (1 << 16) | |
315 | #define BITMAP_SIZE (1 << 13) | |
316 | ||
317 | static uint16_t reverse_lut(const uint8_t *bitmap, uint16_t *lut) | |
318 | { | |
319 | int i, k = 0; | |
320 | ||
321 | for (i = 0; i < USHORT_RANGE; i++) | |
322 | if ((i == 0) || (bitmap[i >> 3] & (1 << (i & 7)))) | |
323 | lut[k++] = i; | |
324 | ||
325 | i = k - 1; | |
326 | ||
327 | memset(lut + k, 0, (USHORT_RANGE - k) * 2); | |
328 | ||
329 | return i; | |
330 | } | |
331 | ||
332 | static void apply_lut(const uint16_t *lut, uint16_t *dst, int dsize) | |
333 | { | |
334 | int i; | |
335 | ||
336 | for (i = 0; i < dsize; ++i) | |
337 | dst[i] = lut[dst[i]]; | |
338 | } | |
339 | ||
340 | #define HUF_ENCBITS 16 // literal (value) bit length | |
341 | #define HUF_DECBITS 14 // decoding bit size (>= 8) | |
342 | ||
343 | #define HUF_ENCSIZE ((1 << HUF_ENCBITS) + 1) // encoding table size | |
344 | #define HUF_DECSIZE (1 << HUF_DECBITS) // decoding table size | |
345 | #define HUF_DECMASK (HUF_DECSIZE - 1) | |
346 | ||
347 | typedef struct HufDec { | |
348 | int len; | |
349 | int lit; | |
350 | int *p; | |
351 | } HufDec; | |
352 | ||
353 | static void huf_canonical_code_table(uint64_t *hcode) | |
354 | { | |
355 | uint64_t c, n[59] = { 0 }; | |
356 | int i; | |
357 | ||
358 | for (i = 0; i < HUF_ENCSIZE; ++i) | |
359 | n[hcode[i]] += 1; | |
360 | ||
361 | c = 0; | |
362 | for (i = 58; i > 0; --i) { | |
363 | uint64_t nc = ((c + n[i]) >> 1); | |
364 | n[i] = c; | |
365 | c = nc; | |
366 | } | |
367 | ||
368 | for (i = 0; i < HUF_ENCSIZE; ++i) { | |
369 | int l = hcode[i]; | |
370 | ||
371 | if (l > 0) | |
372 | hcode[i] = l | (n[l]++ << 6); | |
373 | } | |
374 | } | |
375 | ||
376 | #define SHORT_ZEROCODE_RUN 59 | |
377 | #define LONG_ZEROCODE_RUN 63 | |
378 | #define SHORTEST_LONG_RUN (2 + LONG_ZEROCODE_RUN - SHORT_ZEROCODE_RUN) | |
379 | #define LONGEST_LONG_RUN (255 + SHORTEST_LONG_RUN) | |
380 | ||
381 | static int huf_unpack_enc_table(GetByteContext *gb, | |
382 | int32_t im, int32_t iM, uint64_t *hcode) | |
383 | { | |
384 | GetBitContext gbit; | |
385 | ||
386 | init_get_bits8(&gbit, gb->buffer, bytestream2_get_bytes_left(gb)); | |
387 | ||
388 | for (; im <= iM; im++) { | |
389 | uint64_t l = hcode[im] = get_bits(&gbit, 6); | |
390 | ||
391 | if (l == LONG_ZEROCODE_RUN) { | |
392 | int zerun = get_bits(&gbit, 8) + SHORTEST_LONG_RUN; | |
393 | ||
394 | if (im + zerun > iM + 1) | |
395 | return AVERROR_INVALIDDATA; | |
396 | ||
397 | while (zerun--) | |
398 | hcode[im++] = 0; | |
399 | ||
400 | im--; | |
401 | } else if (l >= SHORT_ZEROCODE_RUN) { | |
402 | int zerun = l - SHORT_ZEROCODE_RUN + 2; | |
403 | ||
404 | if (im + zerun > iM + 1) | |
405 | return AVERROR_INVALIDDATA; | |
406 | ||
407 | while (zerun--) | |
408 | hcode[im++] = 0; | |
409 | ||
410 | im--; | |
411 | } | |
412 | } | |
413 | ||
414 | bytestream2_skip(gb, (get_bits_count(&gbit) + 7) / 8); | |
415 | huf_canonical_code_table(hcode); | |
416 | ||
417 | return 0; | |
418 | } | |
419 | ||
420 | static int huf_build_dec_table(const uint64_t *hcode, int im, | |
421 | int iM, HufDec *hdecod) | |
422 | { | |
423 | for (; im <= iM; im++) { | |
424 | uint64_t c = hcode[im] >> 6; | |
425 | int i, l = hcode[im] & 63; | |
426 | ||
427 | if (c >> l) | |
428 | return AVERROR_INVALIDDATA; | |
429 | ||
430 | if (l > HUF_DECBITS) { | |
431 | HufDec *pl = hdecod + (c >> (l - HUF_DECBITS)); | |
432 | if (pl->len) | |
433 | return AVERROR_INVALIDDATA; | |
434 | ||
435 | pl->lit++; | |
436 | ||
437 | pl->p = av_realloc(pl->p, pl->lit * sizeof(int)); | |
438 | if (!pl->p) | |
439 | return AVERROR(ENOMEM); | |
440 | ||
441 | pl->p[pl->lit - 1] = im; | |
442 | } else if (l) { | |
443 | HufDec *pl = hdecod + (c << (HUF_DECBITS - l)); | |
444 | ||
445 | for (i = 1 << (HUF_DECBITS - l); i > 0; i--, pl++) { | |
446 | if (pl->len || pl->p) | |
447 | return AVERROR_INVALIDDATA; | |
448 | pl->len = l; | |
449 | pl->lit = im; | |
450 | } | |
451 | } | |
452 | } | |
453 | ||
454 | return 0; | |
455 | } | |
456 | ||
457 | #define get_char(c, lc, gb) \ | |
458 | { \ | |
459 | c = (c << 8) | bytestream2_get_byte(gb); \ | |
460 | lc += 8; \ | |
461 | } | |
462 | ||
463 | #define get_code(po, rlc, c, lc, gb, out, oe) \ | |
464 | { \ | |
465 | if (po == rlc) { \ | |
466 | if (lc < 8) \ | |
467 | get_char(c, lc, gb); \ | |
468 | lc -= 8; \ | |
469 | \ | |
470 | cs = c >> lc; \ | |
471 | \ | |
472 | if (out + cs > oe) \ | |
473 | return AVERROR_INVALIDDATA; \ | |
474 | \ | |
475 | s = out[-1]; \ | |
476 | \ | |
477 | while (cs-- > 0) \ | |
478 | *out++ = s; \ | |
479 | } else if (out < oe) { \ | |
480 | *out++ = po; \ | |
481 | } else { \ | |
482 | return AVERROR_INVALIDDATA; \ | |
483 | } \ | |
484 | } | |
485 | ||
486 | static int huf_decode(const uint64_t *hcode, const HufDec *hdecod, | |
487 | GetByteContext *gb, int nbits, | |
488 | int rlc, int no, uint16_t *out) | |
489 | { | |
490 | uint64_t c = 0; | |
491 | uint16_t *outb = out; | |
492 | uint16_t *oe = out + no; | |
493 | const uint8_t *ie = gb->buffer + (nbits + 7) / 8; // input byte size | |
494 | uint8_t cs, s; | |
495 | int i, lc = 0; | |
496 | ||
497 | while (gb->buffer < ie) { | |
498 | get_char(c, lc, gb); | |
499 | ||
500 | while (lc >= HUF_DECBITS) { | |
501 | const HufDec pl = hdecod[(c >> (lc - HUF_DECBITS)) & HUF_DECMASK]; | |
502 | ||
503 | if (pl.len) { | |
504 | lc -= pl.len; | |
505 | get_code(pl.lit, rlc, c, lc, gb, out, oe); | |
506 | } else { | |
507 | int j; | |
508 | ||
509 | if (!pl.p) | |
510 | return AVERROR_INVALIDDATA; | |
511 | ||
512 | for (j = 0; j < pl.lit; j++) { | |
513 | int l = hcode[pl.p[j]] & 63; | |
514 | ||
515 | while (lc < l && bytestream2_get_bytes_left(gb) > 0) | |
516 | get_char(c, lc, gb); | |
517 | ||
518 | if (lc >= l) { | |
519 | if ((hcode[pl.p[j]] >> 6) == | |
520 | ((c >> (lc - l)) & ((1LL << l) - 1))) { | |
521 | lc -= l; | |
522 | get_code(pl.p[j], rlc, c, lc, gb, out, oe); | |
523 | break; | |
524 | } | |
525 | } | |
526 | } | |
527 | ||
528 | if (j == pl.lit) | |
529 | return AVERROR_INVALIDDATA; | |
530 | } | |
531 | } | |
532 | } | |
533 | ||
534 | i = (8 - nbits) & 7; | |
535 | c >>= i; | |
536 | lc -= i; | |
537 | ||
538 | while (lc > 0) { | |
539 | const HufDec pl = hdecod[(c << (HUF_DECBITS - lc)) & HUF_DECMASK]; | |
540 | ||
541 | if (pl.len) { | |
542 | lc -= pl.len; | |
543 | get_code(pl.lit, rlc, c, lc, gb, out, oe); | |
544 | } else { | |
545 | return AVERROR_INVALIDDATA; | |
546 | } | |
547 | } | |
548 | ||
549 | if (out - outb != no) | |
550 | return AVERROR_INVALIDDATA; | |
551 | return 0; | |
552 | } | |
553 | ||
554 | static int huf_uncompress(GetByteContext *gb, | |
555 | uint16_t *dst, int dst_size) | |
556 | { | |
557 | int32_t src_size, im, iM; | |
558 | uint32_t nBits; | |
559 | uint64_t *freq; | |
560 | HufDec *hdec; | |
561 | int ret, i; | |
562 | ||
563 | src_size = bytestream2_get_le32(gb); | |
564 | im = bytestream2_get_le32(gb); | |
565 | iM = bytestream2_get_le32(gb); | |
566 | bytestream2_skip(gb, 4); | |
567 | nBits = bytestream2_get_le32(gb); | |
568 | if (im < 0 || im >= HUF_ENCSIZE || | |
569 | iM < 0 || iM >= HUF_ENCSIZE || | |
570 | src_size < 0) | |
571 | return AVERROR_INVALIDDATA; | |
572 | ||
573 | bytestream2_skip(gb, 4); | |
574 | ||
575 | freq = av_mallocz_array(HUF_ENCSIZE, sizeof(*freq)); | |
576 | hdec = av_mallocz_array(HUF_DECSIZE, sizeof(*hdec)); | |
577 | if (!freq || !hdec) { | |
578 | ret = AVERROR(ENOMEM); | |
579 | goto fail; | |
580 | } | |
581 | ||
582 | if ((ret = huf_unpack_enc_table(gb, im, iM, freq)) < 0) | |
583 | goto fail; | |
584 | ||
585 | if (nBits > 8 * bytestream2_get_bytes_left(gb)) { | |
586 | ret = AVERROR_INVALIDDATA; | |
587 | goto fail; | |
588 | } | |
589 | ||
590 | if ((ret = huf_build_dec_table(freq, im, iM, hdec)) < 0) | |
591 | goto fail; | |
592 | ret = huf_decode(freq, hdec, gb, nBits, iM, dst_size, dst); | |
593 | ||
594 | fail: | |
595 | for (i = 0; i < HUF_DECSIZE; i++) | |
596 | if (hdec) | |
597 | av_freep(&hdec[i].p); | |
598 | ||
599 | av_free(freq); | |
600 | av_free(hdec); | |
601 | ||
602 | return ret; | |
603 | } | |
604 | ||
605 | static inline void wdec14(uint16_t l, uint16_t h, uint16_t *a, uint16_t *b) | |
606 | { | |
607 | int16_t ls = l; | |
608 | int16_t hs = h; | |
609 | int hi = hs; | |
610 | int ai = ls + (hi & 1) + (hi >> 1); | |
611 | int16_t as = ai; | |
612 | int16_t bs = ai - hi; | |
613 | ||
614 | *a = as; | |
615 | *b = bs; | |
616 | } | |
617 | ||
618 | #define NBITS 16 | |
619 | #define A_OFFSET (1 << (NBITS - 1)) | |
620 | #define MOD_MASK ((1 << NBITS) - 1) | |
621 | ||
622 | static inline void wdec16(uint16_t l, uint16_t h, uint16_t *a, uint16_t *b) | |
623 | { | |
624 | int m = l; | |
625 | int d = h; | |
626 | int bb = (m - (d >> 1)) & MOD_MASK; | |
627 | int aa = (d + bb - A_OFFSET) & MOD_MASK; | |
628 | *b = bb; | |
629 | *a = aa; | |
630 | } | |
631 | ||
632 | static void wav_decode(uint16_t *in, int nx, int ox, | |
633 | int ny, int oy, uint16_t mx) | |
634 | { | |
635 | int w14 = (mx < (1 << 14)); | |
636 | int n = (nx > ny) ? ny : nx; | |
637 | int p = 1; | |
638 | int p2; | |
639 | ||
640 | while (p <= n) | |
641 | p <<= 1; | |
642 | ||
643 | p >>= 1; | |
644 | p2 = p; | |
645 | p >>= 1; | |
646 | ||
647 | while (p >= 1) { | |
648 | uint16_t *py = in; | |
649 | uint16_t *ey = in + oy * (ny - p2); | |
650 | uint16_t i00, i01, i10, i11; | |
651 | int oy1 = oy * p; | |
652 | int oy2 = oy * p2; | |
653 | int ox1 = ox * p; | |
654 | int ox2 = ox * p2; | |
655 | ||
656 | for (; py <= ey; py += oy2) { | |
657 | uint16_t *px = py; | |
658 | uint16_t *ex = py + ox * (nx - p2); | |
659 | ||
660 | for (; px <= ex; px += ox2) { | |
661 | uint16_t *p01 = px + ox1; | |
662 | uint16_t *p10 = px + oy1; | |
663 | uint16_t *p11 = p10 + ox1; | |
664 | ||
665 | if (w14) { | |
666 | wdec14(*px, *p10, &i00, &i10); | |
667 | wdec14(*p01, *p11, &i01, &i11); | |
668 | wdec14(i00, i01, px, p01); | |
669 | wdec14(i10, i11, p10, p11); | |
670 | } else { | |
671 | wdec16(*px, *p10, &i00, &i10); | |
672 | wdec16(*p01, *p11, &i01, &i11); | |
673 | wdec16(i00, i01, px, p01); | |
674 | wdec16(i10, i11, p10, p11); | |
675 | } | |
676 | } | |
677 | ||
678 | if (nx & p) { | |
679 | uint16_t *p10 = px + oy1; | |
680 | ||
681 | if (w14) | |
682 | wdec14(*px, *p10, &i00, p10); | |
683 | else | |
684 | wdec16(*px, *p10, &i00, p10); | |
685 | ||
686 | *px = i00; | |
687 | } | |
688 | } | |
689 | ||
690 | if (ny & p) { | |
691 | uint16_t *px = py; | |
692 | uint16_t *ex = py + ox * (nx - p2); | |
693 | ||
694 | for (; px <= ex; px += ox2) { | |
695 | uint16_t *p01 = px + ox1; | |
696 | ||
697 | if (w14) | |
698 | wdec14(*px, *p01, &i00, p01); | |
699 | else | |
700 | wdec16(*px, *p01, &i00, p01); | |
701 | ||
702 | *px = i00; | |
703 | } | |
704 | } | |
705 | ||
706 | p2 = p; | |
707 | p >>= 1; | |
708 | } | |
709 | } | |
710 | ||
711 | static int piz_uncompress(EXRContext *s, const uint8_t *src, int ssize, | |
712 | int dsize, EXRThreadData *td) | |
713 | { | |
714 | GetByteContext gb; | |
715 | uint16_t maxval, min_non_zero, max_non_zero; | |
716 | uint16_t *ptr; | |
717 | uint16_t *tmp = (uint16_t *)td->tmp; | |
718 | uint8_t *out; | |
719 | int ret, i, j; | |
720 | ||
721 | if (!td->bitmap) | |
722 | td->bitmap = av_malloc(BITMAP_SIZE); | |
723 | if (!td->lut) | |
724 | td->lut = av_malloc(1 << 17); | |
725 | if (!td->bitmap || !td->lut) { | |
726 | av_freep(&td->bitmap); | |
727 | av_freep(&td->lut); | |
728 | return AVERROR(ENOMEM); | |
729 | } | |
730 | ||
731 | bytestream2_init(&gb, src, ssize); | |
732 | min_non_zero = bytestream2_get_le16(&gb); | |
733 | max_non_zero = bytestream2_get_le16(&gb); | |
734 | ||
735 | if (max_non_zero >= BITMAP_SIZE) | |
736 | return AVERROR_INVALIDDATA; | |
737 | ||
738 | memset(td->bitmap, 0, FFMIN(min_non_zero, BITMAP_SIZE)); | |
739 | if (min_non_zero <= max_non_zero) | |
740 | bytestream2_get_buffer(&gb, td->bitmap + min_non_zero, | |
741 | max_non_zero - min_non_zero + 1); | |
742 | memset(td->bitmap + max_non_zero, 0, BITMAP_SIZE - max_non_zero); | |
743 | ||
744 | maxval = reverse_lut(td->bitmap, td->lut); | |
745 | ||
746 | ret = huf_uncompress(&gb, tmp, dsize / sizeof(uint16_t)); | |
747 | if (ret) | |
748 | return ret; | |
749 | ||
750 | ptr = tmp; | |
751 | for (i = 0; i < s->nb_channels; i++) { | |
752 | EXRChannel *channel = &s->channels[i]; | |
753 | int size = channel->pixel_type; | |
754 | ||
755 | for (j = 0; j < size; j++) | |
756 | wav_decode(ptr + j, s->xdelta, size, s->ysize, | |
757 | s->xdelta * size, maxval); | |
758 | ptr += s->xdelta * s->ysize * size; | |
759 | } | |
760 | ||
761 | apply_lut(td->lut, tmp, dsize / sizeof(uint16_t)); | |
762 | ||
763 | out = td->uncompressed_data; | |
764 | for (i = 0; i < s->ysize; i++) | |
765 | for (j = 0; j < s->nb_channels; j++) { | |
766 | uint16_t *in = tmp + j * s->xdelta * s->ysize + i * s->xdelta; | |
767 | memcpy(out, in, s->xdelta * 2); | |
768 | out += s->xdelta * 2; | |
769 | } | |
770 | ||
771 | return 0; | |
772 | } | |
773 | ||
774 | static int pxr24_uncompress(EXRContext *s, const uint8_t *src, | |
775 | int compressed_size, int uncompressed_size, | |
776 | EXRThreadData *td) | |
777 | { | |
778 | unsigned long dest_len = uncompressed_size; | |
779 | const uint8_t *in = td->tmp; | |
780 | uint8_t *out; | |
781 | int c, i, j; | |
782 | ||
783 | if (uncompress(td->tmp, &dest_len, src, compressed_size) != Z_OK || | |
784 | dest_len != uncompressed_size) | |
785 | return AVERROR_INVALIDDATA; | |
786 | ||
787 | out = td->uncompressed_data; | |
788 | for (i = 0; i < s->ysize; i++) | |
789 | for (c = 0; c < s->nb_channels; c++) { | |
790 | EXRChannel *channel = &s->channels[c]; | |
791 | const uint8_t *ptr[4]; | |
792 | uint32_t pixel = 0; | |
793 | ||
794 | switch (channel->pixel_type) { | |
795 | case EXR_FLOAT: | |
796 | ptr[0] = in; | |
797 | ptr[1] = ptr[0] + s->xdelta; | |
798 | ptr[2] = ptr[1] + s->xdelta; | |
799 | in = ptr[2] + s->xdelta; | |
800 | ||
801 | for (j = 0; j < s->xdelta; ++j) { | |
802 | uint32_t diff = (*(ptr[0]++) << 24) | | |
803 | (*(ptr[1]++) << 16) | | |
804 | (*(ptr[2]++) << 8); | |
805 | pixel += diff; | |
806 | bytestream_put_le32(&out, pixel); | |
807 | } | |
808 | break; | |
809 | case EXR_HALF: | |
810 | ptr[0] = in; | |
811 | ptr[1] = ptr[0] + s->xdelta; | |
812 | in = ptr[1] + s->xdelta; | |
813 | for (j = 0; j < s->xdelta; j++) { | |
814 | uint32_t diff = (*(ptr[0]++) << 8) | *(ptr[1]++); | |
815 | ||
816 | pixel += diff; | |
817 | bytestream_put_le16(&out, pixel); | |
818 | } | |
819 | break; | |
820 | default: | |
821 | return AVERROR_INVALIDDATA; | |
822 | } | |
823 | } | |
824 | ||
825 | return 0; | |
826 | } | |
827 | ||
828 | static int decode_block(AVCodecContext *avctx, void *tdata, | |
829 | int jobnr, int threadnr) | |
830 | { | |
831 | EXRContext *s = avctx->priv_data; | |
832 | AVFrame *const p = s->picture; | |
833 | EXRThreadData *td = &s->thread_data[threadnr]; | |
834 | const uint8_t *channel_buffer[4] = { 0 }; | |
835 | const uint8_t *buf = s->buf; | |
836 | uint64_t line_offset, uncompressed_size; | |
837 | uint32_t xdelta = s->xdelta; | |
838 | uint16_t *ptr_x; | |
839 | uint8_t *ptr; | |
840 | uint32_t data_size, line; | |
841 | const uint8_t *src; | |
842 | int axmax = (avctx->width - (s->xmax + 1)) * 2 * s->desc->nb_components; | |
843 | int bxmin = s->xmin * 2 * s->desc->nb_components; | |
844 | int i, x, buf_size = s->buf_size; | |
845 | int ret; | |
846 | float one_gamma = 1.0f / s->gamma; | |
847 | ||
848 | line_offset = AV_RL64(s->gb.buffer + jobnr * 8); | |
849 | // Check if the buffer has the required bytes needed from the offset | |
850 | if (line_offset > buf_size - 8) | |
851 | return AVERROR_INVALIDDATA; | |
852 | ||
853 | src = buf + line_offset + 8; | |
854 | line = AV_RL32(src - 8); | |
855 | if (line < s->ymin || line > s->ymax) | |
856 | return AVERROR_INVALIDDATA; | |
857 | ||
858 | data_size = AV_RL32(src - 4); | |
859 | if (data_size <= 0 || data_size > buf_size) | |
860 | return AVERROR_INVALIDDATA; | |
861 | ||
862 | s->ysize = FFMIN(s->scan_lines_per_block, s->ymax - line + 1); | |
863 | uncompressed_size = s->scan_line_size * s->ysize; | |
864 | if ((s->compression == EXR_RAW && (data_size != uncompressed_size || | |
865 | line_offset > buf_size - uncompressed_size)) || | |
866 | (s->compression != EXR_RAW && (data_size > uncompressed_size || | |
867 | line_offset > buf_size - data_size))) { | |
868 | return AVERROR_INVALIDDATA; | |
869 | } | |
870 | ||
871 | if (data_size < uncompressed_size) { | |
872 | av_fast_padded_malloc(&td->uncompressed_data, | |
873 | &td->uncompressed_size, uncompressed_size); | |
874 | av_fast_padded_malloc(&td->tmp, &td->tmp_size, uncompressed_size); | |
875 | if (!td->uncompressed_data || !td->tmp) | |
876 | return AVERROR(ENOMEM); | |
877 | ||
878 | ret = AVERROR_INVALIDDATA; | |
879 | switch (s->compression) { | |
880 | case EXR_ZIP1: | |
881 | case EXR_ZIP16: | |
882 | ret = zip_uncompress(src, data_size, uncompressed_size, td); | |
883 | break; | |
884 | case EXR_PIZ: | |
885 | ret = piz_uncompress(s, src, data_size, uncompressed_size, td); | |
886 | break; | |
887 | case EXR_PXR24: | |
888 | ret = pxr24_uncompress(s, src, data_size, uncompressed_size, td); | |
889 | break; | |
890 | case EXR_RLE: | |
891 | ret = rle_uncompress(src, data_size, uncompressed_size, td); | |
892 | } | |
893 | if (ret < 0) { | |
894 | av_log(avctx, AV_LOG_ERROR, "decode_block() failed.\n"); | |
895 | return ret; | |
896 | } | |
897 | src = td->uncompressed_data; | |
898 | } | |
899 | ||
900 | channel_buffer[0] = src + xdelta * s->channel_offsets[0]; | |
901 | channel_buffer[1] = src + xdelta * s->channel_offsets[1]; | |
902 | channel_buffer[2] = src + xdelta * s->channel_offsets[2]; | |
903 | if (s->channel_offsets[3] >= 0) | |
904 | channel_buffer[3] = src + xdelta * s->channel_offsets[3]; | |
905 | ||
906 | ptr = p->data[0] + line * p->linesize[0]; | |
907 | for (i = 0; | |
908 | i < s->scan_lines_per_block && line + i <= s->ymax; | |
909 | i++, ptr += p->linesize[0]) { | |
910 | const uint8_t *r, *g, *b, *a; | |
911 | ||
912 | r = channel_buffer[0]; | |
913 | g = channel_buffer[1]; | |
914 | b = channel_buffer[2]; | |
915 | if (channel_buffer[3]) | |
916 | a = channel_buffer[3]; | |
917 | ||
918 | ptr_x = (uint16_t *) ptr; | |
919 | ||
920 | // Zero out the start if xmin is not 0 | |
921 | memset(ptr_x, 0, bxmin); | |
922 | ptr_x += s->xmin * s->desc->nb_components; | |
923 | if (s->pixel_type == EXR_FLOAT) { | |
924 | // 32-bit | |
925 | for (x = 0; x < xdelta; x++) { | |
926 | union av_intfloat32 t; | |
927 | t.i = bytestream_get_le32(&r); | |
928 | if ( t.f > 0.0f ) /* avoid negative values */ | |
929 | t.f = powf(t.f, one_gamma); | |
930 | *ptr_x++ = exr_flt2uint(t.i); | |
931 | ||
932 | t.i = bytestream_get_le32(&g); | |
933 | if ( t.f > 0.0f ) | |
934 | t.f = powf(t.f, one_gamma); | |
935 | *ptr_x++ = exr_flt2uint(t.i); | |
936 | ||
937 | t.i = bytestream_get_le32(&b); | |
938 | if ( t.f > 0.0f ) | |
939 | t.f = powf(t.f, one_gamma); | |
940 | *ptr_x++ = exr_flt2uint(t.i); | |
941 | if (channel_buffer[3]) | |
942 | *ptr_x++ = exr_flt2uint(bytestream_get_le32(&a)); | |
943 | } | |
944 | } else { | |
945 | // 16-bit | |
946 | for (x = 0; x < xdelta; x++) { | |
947 | *ptr_x++ = s->gamma_table[bytestream_get_le16(&r)]; | |
948 | *ptr_x++ = s->gamma_table[bytestream_get_le16(&g)]; | |
949 | *ptr_x++ = s->gamma_table[bytestream_get_le16(&b)]; | |
950 | if (channel_buffer[3]) | |
951 | *ptr_x++ = exr_halflt2uint(bytestream_get_le16(&a)); | |
952 | } | |
953 | } | |
954 | ||
955 | // Zero out the end if xmax+1 is not w | |
956 | memset(ptr_x, 0, axmax); | |
957 | ||
958 | channel_buffer[0] += s->scan_line_size; | |
959 | channel_buffer[1] += s->scan_line_size; | |
960 | channel_buffer[2] += s->scan_line_size; | |
961 | if (channel_buffer[3]) | |
962 | channel_buffer[3] += s->scan_line_size; | |
963 | } | |
964 | ||
965 | return 0; | |
966 | } | |
967 | ||
968 | /** | |
969 | * Check if the variable name corresponds to its data type. | |
970 | * | |
971 | * @param s the EXRContext | |
972 | * @param value_name name of the variable to check | |
973 | * @param value_type type of the variable to check | |
974 | * @param minimum_length minimum length of the variable data | |
975 | * | |
976 | * @return bytes to read containing variable data | |
977 | * -1 if variable is not found | |
978 | * 0 if buffer ended prematurely | |
979 | */ | |
980 | static int check_header_variable(EXRContext *s, | |
981 | const char *value_name, | |
982 | const char *value_type, | |
983 | unsigned int minimum_length) | |
984 | { | |
985 | int var_size = -1; | |
986 | ||
987 | if (bytestream2_get_bytes_left(&s->gb) >= minimum_length && | |
988 | !strcmp(s->gb.buffer, value_name)) { | |
989 | // found value_name, jump to value_type (null terminated strings) | |
990 | s->gb.buffer += strlen(value_name) + 1; | |
991 | if (!strcmp(s->gb.buffer, value_type)) { | |
992 | s->gb.buffer += strlen(value_type) + 1; | |
993 | var_size = bytestream2_get_le32(&s->gb); | |
994 | // don't go read past boundaries | |
995 | if (var_size > bytestream2_get_bytes_left(&s->gb)) | |
996 | var_size = 0; | |
997 | } else { | |
998 | // value_type not found, reset the buffer | |
999 | s->gb.buffer -= strlen(value_name) + 1; | |
1000 | av_log(s->avctx, AV_LOG_WARNING, | |
1001 | "Unknown data type %s for header variable %s.\n", | |
1002 | value_type, value_name); | |
1003 | } | |
1004 | } | |
1005 | ||
1006 | return var_size; | |
1007 | } | |
1008 | ||
1009 | static int decode_header(EXRContext *s) | |
1010 | { | |
1011 | int current_channel_offset = 0; | |
1012 | int magic_number, version, flags, i; | |
1013 | ||
1014 | if (bytestream2_get_bytes_left(&s->gb) < 10) { | |
1015 | av_log(s->avctx, AV_LOG_ERROR, "Header too short to parse.\n"); | |
1016 | return AVERROR_INVALIDDATA; | |
1017 | } | |
1018 | ||
1019 | magic_number = bytestream2_get_le32(&s->gb); | |
1020 | if (magic_number != 20000630) { | |
1021 | /* As per documentation of OpenEXR, it is supposed to be | |
1022 | * int 20000630 little-endian */ | |
1023 | av_log(s->avctx, AV_LOG_ERROR, "Wrong magic number %d.\n", magic_number); | |
1024 | return AVERROR_INVALIDDATA; | |
1025 | } | |
1026 | ||
1027 | version = bytestream2_get_byte(&s->gb); | |
1028 | if (version != 2) { | |
1029 | avpriv_report_missing_feature(s->avctx, "Version %d", version); | |
1030 | return AVERROR_PATCHWELCOME; | |
1031 | } | |
1032 | ||
1033 | flags = bytestream2_get_le24(&s->gb); | |
1034 | if (flags & 0x02) { | |
1035 | avpriv_report_missing_feature(s->avctx, "Tile support"); | |
1036 | return AVERROR_PATCHWELCOME; | |
1037 | } | |
1038 | ||
1039 | // Parse the header | |
1040 | while (bytestream2_get_bytes_left(&s->gb) > 0 && *s->gb.buffer) { | |
1041 | int var_size; | |
1042 | if ((var_size = check_header_variable(s, "channels", | |
1043 | "chlist", 38)) >= 0) { | |
1044 | GetByteContext ch_gb; | |
1045 | if (!var_size) | |
1046 | return AVERROR_INVALIDDATA; | |
1047 | ||
1048 | bytestream2_init(&ch_gb, s->gb.buffer, var_size); | |
1049 | ||
1050 | while (bytestream2_get_bytes_left(&ch_gb) >= 19) { | |
1051 | EXRChannel *channel; | |
1052 | enum ExrPixelType current_pixel_type; | |
1053 | int channel_index = -1; | |
1054 | int xsub, ysub; | |
1055 | ||
1056 | if (strcmp(s->layer, "") != 0) { | |
1057 | if (strncmp(ch_gb.buffer, s->layer, strlen(s->layer)) == 0) { | |
1058 | ch_gb.buffer += strlen(s->layer); | |
1059 | if (*ch_gb.buffer == '.') | |
1060 | ch_gb.buffer++; /* skip dot if not given */ | |
1061 | av_log(s->avctx, AV_LOG_INFO, | |
1062 | "Layer %s.%s matched.\n", s->layer, ch_gb.buffer); | |
1063 | } | |
1064 | } | |
1065 | ||
1066 | if (!strcmp(ch_gb.buffer, "R") || | |
1067 | !strcmp(ch_gb.buffer, "X") || | |
1068 | !strcmp(ch_gb.buffer, "U")) | |
1069 | channel_index = 0; | |
1070 | else if (!strcmp(ch_gb.buffer, "G") || | |
1071 | !strcmp(ch_gb.buffer, "Y") || | |
1072 | !strcmp(ch_gb.buffer, "V")) | |
1073 | channel_index = 1; | |
1074 | else if (!strcmp(ch_gb.buffer, "B") || | |
1075 | !strcmp(ch_gb.buffer, "Z") || | |
1076 | !strcmp(ch_gb.buffer, "W")) | |
1077 | channel_index = 2; | |
1078 | else if (!strcmp(ch_gb.buffer, "A")) | |
1079 | channel_index = 3; | |
1080 | else | |
1081 | av_log(s->avctx, AV_LOG_WARNING, | |
1082 | "Unsupported channel %.256s.\n", ch_gb.buffer); | |
1083 | ||
1084 | /* skip until you get a 0 */ | |
1085 | while (bytestream2_get_bytes_left(&ch_gb) > 0 && | |
1086 | bytestream2_get_byte(&ch_gb)) | |
1087 | continue; | |
1088 | ||
1089 | if (bytestream2_get_bytes_left(&ch_gb) < 4) { | |
1090 | av_log(s->avctx, AV_LOG_ERROR, "Incomplete header.\n"); | |
1091 | return AVERROR_INVALIDDATA; | |
1092 | } | |
1093 | ||
1094 | current_pixel_type = bytestream2_get_le32(&ch_gb); | |
1095 | if (current_pixel_type >= EXR_UNKNOWN) { | |
1096 | avpriv_report_missing_feature(s->avctx, | |
1097 | "Pixel type %d.\n", | |
1098 | current_pixel_type); | |
1099 | return AVERROR_PATCHWELCOME; | |
1100 | } | |
1101 | ||
1102 | bytestream2_skip(&ch_gb, 4); | |
1103 | xsub = bytestream2_get_le32(&ch_gb); | |
1104 | ysub = bytestream2_get_le32(&ch_gb); | |
1105 | if (xsub != 1 || ysub != 1) { | |
1106 | avpriv_report_missing_feature(s->avctx, | |
1107 | "Subsampling %dx%d", | |
1108 | xsub, ysub); | |
1109 | return AVERROR_PATCHWELCOME; | |
1110 | } | |
1111 | ||
1112 | if (channel_index >= 0) { | |
1113 | if (s->pixel_type != EXR_UNKNOWN && | |
1114 | s->pixel_type != current_pixel_type) { | |
1115 | av_log(s->avctx, AV_LOG_ERROR, | |
1116 | "RGB channels not of the same depth.\n"); | |
1117 | return AVERROR_INVALIDDATA; | |
1118 | } | |
1119 | s->pixel_type = current_pixel_type; | |
1120 | s->channel_offsets[channel_index] = current_channel_offset; | |
1121 | } | |
1122 | ||
1123 | s->channels = av_realloc(s->channels, | |
1124 | ++s->nb_channels * sizeof(EXRChannel)); | |
1125 | if (!s->channels) | |
1126 | return AVERROR(ENOMEM); | |
1127 | channel = &s->channels[s->nb_channels - 1]; | |
1128 | channel->pixel_type = current_pixel_type; | |
1129 | channel->xsub = xsub; | |
1130 | channel->ysub = ysub; | |
1131 | ||
1132 | current_channel_offset += 1 << current_pixel_type; | |
1133 | } | |
1134 | ||
1135 | /* Check if all channels are set with an offset or if the channels | |
1136 | * are causing an overflow */ | |
1137 | if (FFMIN3(s->channel_offsets[0], | |
1138 | s->channel_offsets[1], | |
1139 | s->channel_offsets[2]) < 0) { | |
1140 | if (s->channel_offsets[0] < 0) | |
1141 | av_log(s->avctx, AV_LOG_ERROR, "Missing red channel.\n"); | |
1142 | if (s->channel_offsets[1] < 0) | |
1143 | av_log(s->avctx, AV_LOG_ERROR, "Missing green channel.\n"); | |
1144 | if (s->channel_offsets[2] < 0) | |
1145 | av_log(s->avctx, AV_LOG_ERROR, "Missing blue channel.\n"); | |
1146 | return AVERROR_INVALIDDATA; | |
1147 | } | |
1148 | ||
1149 | // skip one last byte and update main gb | |
1150 | s->gb.buffer = ch_gb.buffer + 1; | |
1151 | continue; | |
1152 | } else if ((var_size = check_header_variable(s, "dataWindow", "box2i", | |
1153 | 31)) >= 0) { | |
1154 | if (!var_size) | |
1155 | return AVERROR_INVALIDDATA; | |
1156 | ||
1157 | s->xmin = bytestream2_get_le32(&s->gb); | |
1158 | s->ymin = bytestream2_get_le32(&s->gb); | |
1159 | s->xmax = bytestream2_get_le32(&s->gb); | |
1160 | s->ymax = bytestream2_get_le32(&s->gb); | |
1161 | s->xdelta = (s->xmax - s->xmin) + 1; | |
1162 | s->ydelta = (s->ymax - s->ymin) + 1; | |
1163 | ||
1164 | continue; | |
1165 | } else if ((var_size = check_header_variable(s, "displayWindow", | |
1166 | "box2i", 34)) >= 0) { | |
1167 | if (!var_size) | |
1168 | return AVERROR_INVALIDDATA; | |
1169 | ||
1170 | bytestream2_skip(&s->gb, 8); | |
1171 | s->w = bytestream2_get_le32(&s->gb) + 1; | |
1172 | s->h = bytestream2_get_le32(&s->gb) + 1; | |
1173 | ||
1174 | continue; | |
1175 | } else if ((var_size = check_header_variable(s, "lineOrder", | |
1176 | "lineOrder", 25)) >= 0) { | |
1177 | int line_order; | |
1178 | if (!var_size) | |
1179 | return AVERROR_INVALIDDATA; | |
1180 | ||
1181 | line_order = bytestream2_get_byte(&s->gb); | |
1182 | av_log(s->avctx, AV_LOG_DEBUG, "line order: %d.\n", line_order); | |
1183 | if (line_order > 2) { | |
1184 | av_log(s->avctx, AV_LOG_ERROR, "Unknown line order.\n"); | |
1185 | return AVERROR_INVALIDDATA; | |
1186 | } | |
1187 | ||
1188 | continue; | |
1189 | } else if ((var_size = check_header_variable(s, "pixelAspectRatio", | |
1190 | "float", 31)) >= 0) { | |
1191 | if (!var_size) | |
1192 | return AVERROR_INVALIDDATA; | |
1193 | ||
1194 | ff_set_sar(s->avctx, | |
1195 | av_d2q(av_int2float(bytestream2_get_le32(&s->gb)), 255)); | |
1196 | ||
1197 | continue; | |
1198 | } else if ((var_size = check_header_variable(s, "compression", | |
1199 | "compression", 29)) >= 0) { | |
1200 | if (!var_size) | |
1201 | return AVERROR_INVALIDDATA; | |
1202 | ||
1203 | if (s->compression == EXR_UNKN) | |
1204 | s->compression = bytestream2_get_byte(&s->gb); | |
1205 | else | |
1206 | av_log(s->avctx, AV_LOG_WARNING, | |
1207 | "Found more than one compression attribute.\n"); | |
1208 | ||
1209 | continue; | |
1210 | } | |
1211 | ||
1212 | // Check if there are enough bytes for a header | |
1213 | if (bytestream2_get_bytes_left(&s->gb) <= 9) { | |
1214 | av_log(s->avctx, AV_LOG_ERROR, "Incomplete header\n"); | |
1215 | return AVERROR_INVALIDDATA; | |
1216 | } | |
1217 | ||
1218 | // Process unknown variables | |
1219 | for (i = 0; i < 2; i++) // value_name and value_type | |
1220 | while (bytestream2_get_byte(&s->gb) != 0); | |
1221 | ||
1222 | // Skip variable length | |
1223 | bytestream2_skip(&s->gb, bytestream2_get_le32(&s->gb)); | |
1224 | } | |
1225 | ||
1226 | if (s->compression == EXR_UNKN) { | |
1227 | av_log(s->avctx, AV_LOG_ERROR, "Missing compression attribute.\n"); | |
1228 | return AVERROR_INVALIDDATA; | |
1229 | } | |
1230 | s->scan_line_size = s->xdelta * current_channel_offset; | |
1231 | ||
1232 | if (bytestream2_get_bytes_left(&s->gb) <= 0) { | |
1233 | av_log(s->avctx, AV_LOG_ERROR, "Incomplete frame.\n"); | |
1234 | return AVERROR_INVALIDDATA; | |
1235 | } | |
1236 | ||
1237 | // aaand we are done | |
1238 | bytestream2_skip(&s->gb, 1); | |
1239 | return 0; | |
1240 | } | |
1241 | ||
1242 | static int decode_frame(AVCodecContext *avctx, void *data, | |
1243 | int *got_frame, AVPacket *avpkt) | |
1244 | { | |
1245 | EXRContext *s = avctx->priv_data; | |
1246 | ThreadFrame frame = { .f = data }; | |
1247 | AVFrame *picture = data; | |
1248 | uint8_t *ptr; | |
1249 | ||
1250 | int y, ret; | |
1251 | int out_line_size; | |
1252 | int scan_line_blocks; | |
1253 | ||
1254 | bytestream2_init(&s->gb, avpkt->data, avpkt->size); | |
1255 | ||
1256 | if ((ret = decode_header(s)) < 0) | |
1257 | return ret; | |
1258 | ||
1259 | switch (s->pixel_type) { | |
1260 | case EXR_FLOAT: | |
1261 | case EXR_HALF: | |
1262 | if (s->channel_offsets[3] >= 0) | |
1263 | avctx->pix_fmt = AV_PIX_FMT_RGBA64; | |
1264 | else | |
1265 | avctx->pix_fmt = AV_PIX_FMT_RGB48; | |
1266 | break; | |
1267 | case EXR_UINT: | |
1268 | avpriv_request_sample(avctx, "32-bit unsigned int"); | |
1269 | return AVERROR_PATCHWELCOME; | |
1270 | default: | |
1271 | av_log(avctx, AV_LOG_ERROR, "Missing channel list.\n"); | |
1272 | return AVERROR_INVALIDDATA; | |
1273 | } | |
1274 | ||
1275 | switch (s->compression) { | |
1276 | case EXR_RAW: | |
1277 | case EXR_RLE: | |
1278 | case EXR_ZIP1: | |
1279 | s->scan_lines_per_block = 1; | |
1280 | break; | |
1281 | case EXR_PXR24: | |
1282 | case EXR_ZIP16: | |
1283 | s->scan_lines_per_block = 16; | |
1284 | break; | |
1285 | case EXR_PIZ: | |
1286 | s->scan_lines_per_block = 32; | |
1287 | break; | |
1288 | default: | |
1289 | avpriv_report_missing_feature(avctx, "Compression %d", s->compression); | |
1290 | return AVERROR_PATCHWELCOME; | |
1291 | } | |
1292 | ||
1293 | /* Verify the xmin, xmax, ymin, ymax and xdelta before setting | |
1294 | * the actual image size. */ | |
1295 | if (s->xmin > s->xmax || | |
1296 | s->ymin > s->ymax || | |
1297 | s->xdelta != s->xmax - s->xmin + 1 || | |
1298 | s->xmax >= s->w || | |
1299 | s->ymax >= s->h) { | |
1300 | av_log(avctx, AV_LOG_ERROR, "Wrong or missing size information.\n"); | |
1301 | return AVERROR_INVALIDDATA; | |
1302 | } | |
1303 | ||
1304 | if ((ret = ff_set_dimensions(avctx, s->w, s->h)) < 0) | |
1305 | return ret; | |
1306 | ||
1307 | s->desc = av_pix_fmt_desc_get(avctx->pix_fmt); | |
1308 | if (!s->desc) | |
1309 | return AVERROR_INVALIDDATA; | |
1310 | out_line_size = avctx->width * 2 * s->desc->nb_components; | |
1311 | scan_line_blocks = (s->ydelta + s->scan_lines_per_block - 1) / | |
1312 | s->scan_lines_per_block; | |
1313 | ||
1314 | if ((ret = ff_thread_get_buffer(avctx, &frame, 0)) < 0) | |
1315 | return ret; | |
1316 | ||
1317 | if (bytestream2_get_bytes_left(&s->gb) < scan_line_blocks * 8) | |
1318 | return AVERROR_INVALIDDATA; | |
1319 | ||
1320 | // save pointer we are going to use in decode_block | |
1321 | s->buf = avpkt->data; | |
1322 | s->buf_size = avpkt->size; | |
1323 | ptr = picture->data[0]; | |
1324 | ||
1325 | // Zero out the start if ymin is not 0 | |
1326 | for (y = 0; y < s->ymin; y++) { | |
1327 | memset(ptr, 0, out_line_size); | |
1328 | ptr += picture->linesize[0]; | |
1329 | } | |
1330 | ||
1331 | s->picture = picture; | |
1332 | avctx->execute2(avctx, decode_block, s->thread_data, NULL, scan_line_blocks); | |
1333 | ||
1334 | // Zero out the end if ymax+1 is not h | |
1335 | for (y = s->ymax + 1; y < avctx->height; y++) { | |
1336 | memset(ptr, 0, out_line_size); | |
1337 | ptr += picture->linesize[0]; | |
1338 | } | |
1339 | ||
1340 | picture->pict_type = AV_PICTURE_TYPE_I; | |
1341 | *got_frame = 1; | |
1342 | ||
1343 | return avpkt->size; | |
1344 | } | |
1345 | ||
1346 | static av_cold int decode_init(AVCodecContext *avctx) | |
1347 | { | |
1348 | uint32_t i; | |
1349 | union av_intfloat32 t; | |
1350 | EXRContext *s = avctx->priv_data; | |
1351 | float one_gamma = 1.0f / s->gamma; | |
1352 | ||
1353 | s->avctx = avctx; | |
1354 | s->xmin = ~0; | |
1355 | s->xmax = ~0; | |
1356 | s->ymin = ~0; | |
1357 | s->ymax = ~0; | |
1358 | s->xdelta = ~0; | |
1359 | s->ydelta = ~0; | |
1360 | s->channel_offsets[0] = -1; | |
1361 | s->channel_offsets[1] = -1; | |
1362 | s->channel_offsets[2] = -1; | |
1363 | s->channel_offsets[3] = -1; | |
1364 | s->pixel_type = EXR_UNKNOWN; | |
1365 | s->compression = EXR_UNKN; | |
1366 | s->nb_channels = 0; | |
1367 | s->w = 0; | |
1368 | s->h = 0; | |
1369 | ||
1370 | if ( one_gamma > 0.9999f && one_gamma < 1.0001f ) { | |
1371 | for ( i = 0; i < 65536; ++i ) { | |
1372 | s->gamma_table[i] = exr_halflt2uint(i); | |
1373 | } | |
1374 | } else { | |
1375 | for ( i = 0; i < 65536; ++i ) { | |
1376 | t = exr_half2float(i); | |
1377 | /* If negative value we reuse half value */ | |
1378 | if ( t.f <= 0.0f ) { | |
1379 | s->gamma_table[i] = exr_halflt2uint(i); | |
1380 | } else { | |
1381 | t.f = powf(t.f, one_gamma); | |
1382 | s->gamma_table[i] = exr_flt2uint(t.i); | |
1383 | } | |
1384 | } | |
1385 | } | |
1386 | ||
1387 | // allocate thread data, used for non EXR_RAW compreesion types | |
1388 | s->thread_data = av_mallocz_array(avctx->thread_count, sizeof(EXRThreadData)); | |
1389 | if (!s->thread_data) | |
1390 | return AVERROR_INVALIDDATA; | |
1391 | ||
1392 | return 0; | |
1393 | } | |
1394 | ||
1395 | static int decode_init_thread_copy(AVCodecContext *avctx) | |
1396 | { EXRContext *s = avctx->priv_data; | |
1397 | ||
1398 | // allocate thread data, used for non EXR_RAW compreesion types | |
1399 | s->thread_data = av_mallocz_array(avctx->thread_count, sizeof(EXRThreadData)); | |
1400 | if (!s->thread_data) | |
1401 | return AVERROR_INVALIDDATA; | |
1402 | ||
1403 | return 0; | |
1404 | } | |
1405 | ||
1406 | static av_cold int decode_end(AVCodecContext *avctx) | |
1407 | { | |
1408 | EXRContext *s = avctx->priv_data; | |
1409 | int i; | |
1410 | for (i = 0; i < avctx->thread_count; i++) { | |
1411 | EXRThreadData *td = &s->thread_data[i]; | |
1412 | av_freep(&td->uncompressed_data); | |
1413 | av_freep(&td->tmp); | |
1414 | av_freep(&td->bitmap); | |
1415 | av_freep(&td->lut); | |
1416 | } | |
1417 | ||
1418 | av_freep(&s->thread_data); | |
1419 | av_freep(&s->channels); | |
1420 | ||
1421 | return 0; | |
1422 | } | |
1423 | ||
1424 | #define OFFSET(x) offsetof(EXRContext, x) | |
1425 | #define VD AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_DECODING_PARAM | |
1426 | static const AVOption options[] = { | |
1427 | { "layer", "Set the decoding layer", OFFSET(layer), | |
1428 | AV_OPT_TYPE_STRING, { .str = "" }, 0, 0, VD }, | |
1429 | { "gamma", "Set the float gamma value when decoding (experimental/unsupported)", OFFSET(gamma), | |
1430 | AV_OPT_TYPE_FLOAT, { .dbl = 1.0f }, 0.001, FLT_MAX, VD }, | |
1431 | { NULL }, | |
1432 | }; | |
1433 | ||
1434 | static const AVClass exr_class = { | |
1435 | .class_name = "EXR", | |
1436 | .item_name = av_default_item_name, | |
1437 | .option = options, | |
1438 | .version = LIBAVUTIL_VERSION_INT, | |
1439 | }; | |
1440 | ||
1441 | AVCodec ff_exr_decoder = { | |
1442 | .name = "exr", | |
1443 | .long_name = NULL_IF_CONFIG_SMALL("OpenEXR image"), | |
1444 | .type = AVMEDIA_TYPE_VIDEO, | |
1445 | .id = AV_CODEC_ID_EXR, | |
1446 | .priv_data_size = sizeof(EXRContext), | |
1447 | .init = decode_init, | |
1448 | .init_thread_copy = ONLY_IF_THREADS_ENABLED(decode_init_thread_copy), | |
1449 | .close = decode_end, | |
1450 | .decode = decode_frame, | |
1451 | .capabilities = CODEC_CAP_DR1 | CODEC_CAP_FRAME_THREADS | | |
1452 | CODEC_CAP_SLICE_THREADS, | |
1453 | .priv_class = &exr_class, | |
1454 | }; |