1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
1720
1721
1722
1723
1724
1725
1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
1738
1739
1740
1741
1742
1743
1744
1745
1746
1747
1748
1749
1750
1751
1752
1753
1754
1755
1756
1757
1758
1759
1760
1761
1762
1763
1764
1765
1766
1767
1768
1769
1770
1771
1772
1773
1774
1775
1776
1777
1778
1779
1780
1781
1782
1783
1784
1785
1786
1787
1788
1789
1790
1791
1792
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
1819
1820
1821
1822
1823
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849
1850
1851
1852
1853
1854
1855
1856
1857
1858
1859
1860
1861
1862
1863
1864
1865
1866
1867
1868
1869
1870
1871
1872
1873
1874
1875
1876
1877
1878
1879
1880
1881
1882
1883
1884
1885
1886
1887
1888
1889
1890
1891
1892
1893
1894
1895
1896
1897
1898
1899
1900
1901
1902
1903
1904
1905
1906
1907
1908
1909
1910
1911
1912
1913
1914
1915
1916
1917
1918
1919
1920
1921
1922
1923
1924
1925
1926
1927
1928
1929
1930
1931
1932
1933
1934
1935
1936
1937
1938
1939
1940
1941
1942
1943
1944
1945
1946
1947
1948
1949
1950
1951
1952
1953
1954
1955
1956
1957
|
/*
* Copyright (c) 2020, the SerenityOS developers.
* Copyright (c) 2022-2023, Lucas Chollet <lucas.chollet@serenityos.org>
*
* SPDX-License-Identifier: BSD-2-Clause
*/
#include <AK/Debug.h>
#include <AK/Endian.h>
#include <AK/Error.h>
#include <AK/FixedArray.h>
#include <AK/HashMap.h>
#include <AK/Math.h>
#include <AK/MemoryStream.h>
#include <AK/NumericLimits.h>
#include <AK/String.h>
#include <AK/Try.h>
#include <AK/Vector.h>
#include <LibGfx/ImageFormats/JPEGLoader.h>
#define JPEG_INVALID 0X0000
// These names are defined in B.1.1.3 - Marker assignments
#define JPEG_APPN0 0XFFE0
#define JPEG_APPN1 0XFFE1
#define JPEG_APPN2 0XFFE2
#define JPEG_APPN3 0XFFE3
#define JPEG_APPN4 0XFFE4
#define JPEG_APPN5 0XFFE5
#define JPEG_APPN6 0XFFE6
#define JPEG_APPN7 0XFFE7
#define JPEG_APPN8 0XFFE8
#define JPEG_APPN9 0XFFE9
#define JPEG_APPN10 0XFFEA
#define JPEG_APPN11 0XFFEB
#define JPEG_APPN12 0XFFEC
#define JPEG_APPN13 0XFFED
#define JPEG_APPN14 0xFFEE
#define JPEG_APPN15 0xFFEF
#define JPEG_RESERVED1 0xFFF1
#define JPEG_RESERVED2 0xFFF2
#define JPEG_RESERVED3 0xFFF3
#define JPEG_RESERVED4 0xFFF4
#define JPEG_RESERVED5 0xFFF5
#define JPEG_RESERVED6 0xFFF6
#define JPEG_RESERVED7 0xFFF7
#define JPEG_RESERVED8 0xFFF8
#define JPEG_RESERVED9 0xFFF9
#define JPEG_RESERVEDA 0xFFFA
#define JPEG_RESERVEDB 0xFFFB
#define JPEG_RESERVEDC 0xFFFC
#define JPEG_RESERVEDD 0xFFFD
#define JPEG_RST0 0xFFD0
#define JPEG_RST1 0xFFD1
#define JPEG_RST2 0xFFD2
#define JPEG_RST3 0xFFD3
#define JPEG_RST4 0xFFD4
#define JPEG_RST5 0xFFD5
#define JPEG_RST6 0xFFD6
#define JPEG_RST7 0xFFD7
#define JPEG_ZRL 0xF0
#define JPEG_DHP 0xFFDE
#define JPEG_EXP 0xFFDF
#define JPEG_DAC 0XFFCC
#define JPEG_DHT 0XFFC4
#define JPEG_DQT 0XFFDB
#define JPEG_EOI 0xFFD9
#define JPEG_DRI 0XFFDD
#define JPEG_SOF0 0XFFC0
#define JPEG_SOF1 0XFFC1
#define JPEG_SOF2 0xFFC2
#define JPEG_SOF15 0xFFCF
#define JPEG_SOI 0XFFD8
#define JPEG_SOS 0XFFDA
#define JPEG_COM 0xFFFE
namespace Gfx {
constexpr static u8 zigzag_map[64] {
0, 1, 8, 16, 9, 2, 3, 10,
17, 24, 32, 25, 18, 11, 4, 5,
12, 19, 26, 33, 40, 48, 41, 34,
27, 20, 13, 6, 7, 14, 21, 28,
35, 42, 49, 56, 57, 50, 43, 36,
29, 22, 15, 23, 30, 37, 44, 51,
58, 59, 52, 45, 38, 31, 39, 46,
53, 60, 61, 54, 47, 55, 62, 63
};
using Marker = u16;
/**
* MCU means group of data units that are coded together. A data unit is an 8x8
* block of component data. In interleaved scans, number of non-interleaved data
* units of a component C is Ch * Cv, where Ch and Cv represent the horizontal &
* vertical subsampling factors of the component, respectively. A MacroBlock is
* an 8x8 block of RGB values before encoding, and 8x8 block of YCbCr values when
* we're done decoding the huffman stream.
*/
struct Macroblock {
union {
i16 y[64] = { 0 };
i16 r[64];
};
union {
i16 cb[64] = { 0 };
i16 g[64];
};
union {
i16 cr[64] = { 0 };
i16 b[64];
};
i16 k[64] = { 0 };
};
struct MacroblockMeta {
u32 total { 0 };
u32 padded_total { 0 };
u32 hcount { 0 };
u32 vcount { 0 };
u32 hpadded_count { 0 };
u32 vpadded_count { 0 };
};
// In the JPEG format, components are defined first at the frame level, then
// referenced in each scan and aggregated with scan-specific information. The
// two following structs mimic this hierarchy.
struct Component {
// B.2.2 - Frame header syntax
u8 id { 0 }; // Ci, Component identifier
u8 hsample_factor { 1 }; // Hi, Horizontal sampling factor
u8 vsample_factor { 1 }; // Vi, Vertical sampling factor
u8 quantization_table_id { 0 }; // Tqi, Quantization table destination selector
// The JPEG specification does not specify which component corresponds to
// Y, Cb or Cr. This field (actually the index in the parent Vector) will
// act as an authority to determine the *real* component.
// Please note that this is implementation specific.
u8 index { 0 };
};
struct ScanComponent {
// B.2.3 - Scan header syntax
Component& component;
u8 dc_destination_id { 0 }; // Tdj, DC entropy coding table destination selector
u8 ac_destination_id { 0 }; // Taj, AC entropy coding table destination selector
};
struct StartOfFrame {
// Of these, only the first 3 are in mainstream use, and refers to SOF0-2.
enum class FrameType {
Baseline_DCT = 0,
Extended_Sequential_DCT = 1,
Progressive_DCT = 2,
Sequential_Lossless = 3,
Differential_Sequential_DCT = 5,
Differential_Progressive_DCT = 6,
Differential_Sequential_Lossless = 7,
Extended_Sequential_DCT_Arithmetic = 9,
Progressive_DCT_Arithmetic = 10,
Sequential_Lossless_Arithmetic = 11,
Differential_Sequential_DCT_Arithmetic = 13,
Differential_Progressive_DCT_Arithmetic = 14,
Differential_Sequential_Lossless_Arithmetic = 15,
};
FrameType type { FrameType::Baseline_DCT };
u8 precision { 0 };
u16 height { 0 };
u16 width { 0 };
};
struct HuffmanTable {
u8 type { 0 };
u8 destination_id { 0 };
u8 code_counts[16] = { 0 };
Vector<u8> symbols;
Vector<u16> codes;
// Note: The value 8 is chosen quite arbitrarily, the only current constraint
// is that both the symbol and the size fit in an u16. I've tested more
// values but none stand out, and 8 is the value used by libjpeg-turbo.
static constexpr u8 bits_per_cached_code = 8;
static constexpr u8 maximum_bits_per_code = 16;
u8 first_non_cached_code_index {};
void generate_codes()
{
unsigned code = 0;
for (auto number_of_codes : code_counts) {
for (int i = 0; i < number_of_codes; i++)
codes.append(code++);
code <<= 1;
}
generate_lookup_table();
}
struct SymbolAndSize {
u8 symbol {};
u8 size {};
};
ErrorOr<SymbolAndSize> symbol_from_code(u16 code) const
{
static constexpr u8 shift_for_cache = maximum_bits_per_code - bits_per_cached_code;
if (lookup_table[code >> shift_for_cache] != invalid_entry) {
u8 const code_length = lookup_table[code >> shift_for_cache] >> bits_per_cached_code;
return SymbolAndSize { static_cast<u8>(lookup_table[code >> shift_for_cache]), code_length };
}
u64 code_cursor = first_non_cached_code_index;
for (u8 i = HuffmanTable::bits_per_cached_code; i < 16; i++) {
auto const result = code >> (maximum_bits_per_code - 1 - i);
for (u32 j = 0; j < code_counts[i]; j++) {
if (result == codes[code_cursor])
return SymbolAndSize { symbols[code_cursor], static_cast<u8>(i + 1) };
code_cursor++;
}
}
return Error::from_string_literal("This kind of JPEG is not yet supported by the decoder");
}
private:
static constexpr u16 invalid_entry = 0xFF;
void generate_lookup_table()
{
lookup_table.fill(invalid_entry);
u32 code_offset = 0;
for (u8 code_length = 1; code_length <= bits_per_cached_code; code_length++) {
for (u32 i = 0; i < code_counts[code_length - 1]; i++, code_offset++) {
u32 code_key = codes[code_offset] << (bits_per_cached_code - code_length);
for (u8 duplicate_count = 1 << (bits_per_cached_code - code_length); duplicate_count > 0; duplicate_count--) {
lookup_table[code_key] = (code_length << bits_per_cached_code) | symbols[code_offset];
code_key++;
}
}
}
}
Array<u16, 1 << bits_per_cached_code> lookup_table {};
};
class HuffmanStream {
public:
static ErrorOr<HuffmanStream> create(SeekableStream& stream)
{
HuffmanStream huffman {};
u8 last_byte {};
u8 current_byte = TRY(stream.read_value<u8>());
for (;;) {
last_byte = current_byte;
current_byte = TRY(stream.read_value<u8>());
if (last_byte == 0xFF) {
if (current_byte == 0xFF)
continue;
if (current_byte == 0x00) {
current_byte = TRY(stream.read_value<u8>());
huffman.m_stream.append(last_byte);
continue;
}
Marker marker = 0xFF00 | current_byte;
if (marker >= JPEG_RST0 && marker <= JPEG_RST7) {
huffman.m_stream.append(marker);
current_byte = TRY(stream.read_value<u8>());
continue;
}
// Rollback the marker we just read
TRY(stream.seek(-2, AK::SeekMode::FromCurrentPosition));
return huffman;
}
huffman.m_stream.append(last_byte);
}
VERIFY_NOT_REACHED();
}
ErrorOr<u8> next_symbol(HuffmanTable const& table)
{
u16 const code = peek_bits(HuffmanTable::maximum_bits_per_code);
auto const symbol_and_size = TRY(table.symbol_from_code(code));
discard_bits(symbol_and_size.size);
return symbol_and_size.symbol;
}
ErrorOr<u16> read_bits(u8 count = 1)
{
if (count > NumericLimits<u16>::digits()) {
dbgln_if(JPEG_DEBUG, "Can't read {} bits at once!", count);
return Error::from_string_literal("Reading too much huffman bits at once");
}
u16 const value = peek_bits(count);
discard_bits(count);
return value;
}
u16 peek_bits(u8 count) const
{
using BufferType = u32;
constexpr static auto max = NumericLimits<BufferType>::max();
auto const mask = max >> (8 + m_bit_offset);
BufferType msb_buffer {};
if (m_byte_offset + 0 < m_stream.size())
msb_buffer |= (static_cast<BufferType>(m_stream[m_byte_offset + 0]) << (2 * 8));
if (m_byte_offset + 1 < m_stream.size())
msb_buffer |= (static_cast<BufferType>(m_stream[m_byte_offset + 1]) << (1 * 8));
if (m_byte_offset + 2 < m_stream.size())
msb_buffer |= (static_cast<BufferType>(m_stream[m_byte_offset + 2]) << (0 * 8));
return (mask & msb_buffer) >> (3 * 8 - m_bit_offset - count);
}
void discard_bits(u8 count)
{
m_bit_offset += count;
auto const carry = m_bit_offset / 8;
m_bit_offset -= 8 * carry;
m_byte_offset += carry;
}
void advance_to_byte_boundary()
{
if (m_bit_offset > 0) {
m_bit_offset = 0;
m_byte_offset++;
}
}
void skip_byte()
{
m_byte_offset++;
}
u64 byte_offset() const
{
return m_byte_offset;
}
private:
Vector<u8> m_stream;
u8 m_bit_offset { 0 };
u64 m_byte_offset { 0 };
};
struct ICCMultiChunkState {
u8 seen_number_of_icc_chunks { 0 };
FixedArray<ByteBuffer> chunks;
};
struct Scan {
// B.2.3 - Scan header syntax
Vector<ScanComponent, 4> components;
u8 spectral_selection_start {}; // Ss
u8 spectral_selection_end {}; // Se
u8 successive_approximation_high {}; // Ah
u8 successive_approximation_low {}; // Al
HuffmanStream huffman_stream;
u64 end_of_bands_run_count { 0 };
// See the note on Figure B.4 - Scan header syntax
bool are_components_interleaved() const
{
return components.size() != 1;
}
};
enum class ColorTransform {
// https://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-T.872-201206-I!!PDF-E&type=items
// 6.5.3 - APP14 marker segment for colour encoding
CmykOrRgb = 0,
YCbCr = 1,
YCCK = 2,
};
struct JPEGLoadingContext {
enum State {
NotDecoded = 0,
Error,
FrameDecoded,
HeaderDecoded,
BitmapDecoded
};
State state { State::NotDecoded };
Array<Optional<Array<u16, 64>>, 4> quantization_tables {};
StartOfFrame frame;
u8 hsample_factor { 0 };
u8 vsample_factor { 0 };
Scan current_scan;
Vector<Component, 4> components;
RefPtr<Gfx::Bitmap> bitmap;
u16 dc_restart_interval { 0 };
HashMap<u8, HuffmanTable> dc_tables;
HashMap<u8, HuffmanTable> ac_tables;
Array<i16, 4> previous_dc_values {};
MacroblockMeta mblock_meta;
OwnPtr<FixedMemoryStream> stream;
Optional<ColorTransform> color_transform {};
Optional<ICCMultiChunkState> icc_multi_chunk_state;
Optional<ByteBuffer> icc_data;
};
static inline auto* get_component(Macroblock& block, unsigned component)
{
switch (component) {
case 0:
return block.y;
case 1:
return block.cb;
case 2:
return block.cr;
case 3:
return block.k;
default:
VERIFY_NOT_REACHED();
}
}
static ErrorOr<void> refine_coefficient(Scan& scan, auto& coefficient)
{
// G.1.2.3 - Coding model for subsequent scans of successive approximation
// See the correction bit from rule b.
u8 const bit = TRY(scan.huffman_stream.read_bits(1));
if (bit == 1)
coefficient |= 1 << scan.successive_approximation_low;
return {};
}
static ErrorOr<void> add_dc(JPEGLoadingContext& context, Macroblock& macroblock, ScanComponent const& scan_component)
{
auto maybe_table = context.dc_tables.get(scan_component.dc_destination_id);
if (!maybe_table.has_value()) {
dbgln_if(JPEG_DEBUG, "Unable to find a DC table with id: {}", scan_component.dc_destination_id);
return Error::from_string_literal("Unable to find corresponding DC table");
}
auto& dc_table = maybe_table.value();
auto& scan = context.current_scan;
auto* select_component = get_component(macroblock, scan_component.component.index);
auto& coefficient = select_component[0];
if (context.current_scan.successive_approximation_high > 0) {
TRY(refine_coefficient(scan, coefficient));
return {};
}
// For DC coefficients, symbol encodes the length of the coefficient.
auto dc_length = TRY(scan.huffman_stream.next_symbol(dc_table));
// F.1.2.1.2 - Defining Huffman tables for the DC coefficients
// F.1.5.1 - Structure of DC code table for 12-bit sample precision
if ((context.frame.precision == 8 && dc_length > 11)
|| (context.frame.precision == 12 && dc_length > 15)) {
dbgln_if(JPEG_DEBUG, "DC coefficient too long: {}!", dc_length);
return Error::from_string_literal("DC coefficient too long");
}
// DC coefficients are encoded as the difference between previous and current DC values.
i16 dc_diff = TRY(scan.huffman_stream.read_bits(dc_length));
// If MSB in diff is 0, the difference is -ve. Otherwise +ve.
if (dc_length != 0 && dc_diff < (1 << (dc_length - 1)))
dc_diff -= (1 << dc_length) - 1;
auto& previous_dc = context.previous_dc_values[scan_component.component.index];
previous_dc += dc_diff;
coefficient = previous_dc << scan.successive_approximation_low;
return {};
}
static ErrorOr<bool> read_eob(Scan& scan, u32 symbol)
{
// G.1.2.2 - Progressive encoding of AC coefficients with Huffman coding
// Note: We also use it for non-progressive encoding as it supports both EOB and ZRL
if (auto const eob = symbol & 0x0F; eob == 0 && symbol != JPEG_ZRL) {
// We encountered an EOB marker
auto const eob_base = symbol >> 4;
auto const additional_value = TRY(scan.huffman_stream.read_bits(eob_base));
scan.end_of_bands_run_count = additional_value + (1 << eob_base) - 1;
// end_of_bands_run_count is decremented at the end of `build_macroblocks`.
// And we need to now that we reached End of Block in `add_ac`.
++scan.end_of_bands_run_count;
return true;
}
return false;
}
static bool is_progressive(StartOfFrame::FrameType frame_type)
{
return frame_type == StartOfFrame::FrameType::Progressive_DCT
|| frame_type == StartOfFrame::FrameType::Progressive_DCT_Arithmetic
|| frame_type == StartOfFrame::FrameType::Differential_Progressive_DCT
|| frame_type == StartOfFrame::FrameType::Differential_Progressive_DCT_Arithmetic;
}
static ErrorOr<void> add_ac(JPEGLoadingContext& context, Macroblock& macroblock, ScanComponent const& scan_component)
{
auto maybe_table = context.ac_tables.get(scan_component.ac_destination_id);
if (!maybe_table.has_value()) {
dbgln_if(JPEG_DEBUG, "Unable to find a AC table with id: {}", scan_component.ac_destination_id);
return Error::from_string_literal("Unable to find corresponding AC table");
}
auto& ac_table = maybe_table.value();
auto* select_component = get_component(macroblock, scan_component.component.index);
auto& scan = context.current_scan;
// Compute the AC coefficients.
// 0th coefficient is the dc, which is already handled
auto first_coefficient = max(1, scan.spectral_selection_start);
u32 to_skip = 0;
Optional<u8> saved_symbol;
Optional<u8> saved_bit_for_rule_a;
bool in_zrl = false;
for (int j = first_coefficient; j <= scan.spectral_selection_end; ++j) {
auto& coefficient = select_component[zigzag_map[j]];
// AC symbols encode 2 pieces of information, the high 4 bits represent
// number of zeroes to be stuffed before reading the coefficient. Low 4
// bits represent the magnitude of the coefficient.
if (!in_zrl && scan.end_of_bands_run_count == 0 && !saved_symbol.has_value()) {
saved_symbol = TRY(scan.huffman_stream.next_symbol(ac_table));
if (!TRY(read_eob(scan, *saved_symbol))) {
to_skip = *saved_symbol >> 4;
in_zrl = *saved_symbol == JPEG_ZRL;
if (in_zrl) {
to_skip++;
saved_symbol.clear();
}
if (!in_zrl && is_progressive(context.frame.type) && scan.successive_approximation_high != 0) {
// G.1.2.3 - Coding model for subsequent scans of successive approximation
// Bit sign from rule a
saved_bit_for_rule_a = TRY(scan.huffman_stream.read_bits(1));
}
}
}
if (coefficient != 0) {
TRY(refine_coefficient(scan, coefficient));
continue;
}
if (to_skip > 0) {
--to_skip;
if (to_skip == 0)
in_zrl = false;
continue;
}
if (scan.end_of_bands_run_count > 0)
continue;
if (is_progressive(context.frame.type) && scan.successive_approximation_high != 0) {
// G.1.2.3 - Coding model for subsequent scans of successive approximation
if (auto const low_bits = *saved_symbol & 0x0F; low_bits != 1) {
dbgln_if(JPEG_DEBUG, "AC coefficient low bits isn't equal to 1: {}!", low_bits);
return Error::from_string_literal("AC coefficient low bits isn't equal to 1");
}
coefficient = (*saved_bit_for_rule_a == 0 ? -1 : 1) << scan.successive_approximation_low;
saved_bit_for_rule_a.clear();
} else {
// F.1.2.2 - Huffman encoding of AC coefficients
u8 const coeff_length = *saved_symbol & 0x0F;
// F.1.2.2.1 - Structure of AC code table
// F.1.5.2 - Structure of AC code table for 12-bit sample precision
if ((context.frame.precision == 8 && coeff_length > 10)
|| (context.frame.precision == 12 && coeff_length > 14)) {
dbgln_if(JPEG_DEBUG, "AC coefficient too long: {}!", coeff_length);
return Error::from_string_literal("AC coefficient too long");
}
if (coeff_length != 0) {
i32 ac_coefficient = TRY(scan.huffman_stream.read_bits(coeff_length));
if (ac_coefficient < (1 << (coeff_length - 1)))
ac_coefficient -= (1 << coeff_length) - 1;
coefficient = ac_coefficient * (1 << scan.successive_approximation_low);
}
}
saved_symbol.clear();
}
if (to_skip > 0) {
dbgln_if(JPEG_DEBUG, "Run-length exceeded boundaries. Cursor: {}, Skipping: {}!", scan.spectral_selection_end + to_skip, to_skip);
return Error::from_string_literal("Run-length exceeded boundaries");
}
return {};
}
/**
* Build the macroblocks possible by reading single (MCU) subsampled pair of CbCr.
* Depending on the sampling factors, we may not see triples of y, cb, cr in that
* order. If sample factors differ from one, we'll read more than one block of y-
* coefficients before we get to read a cb-cr block.
* In the function below, `hcursor` and `vcursor` denote the location of the block
* we're building in the macroblock matrix. `vfactor_i` and `hfactor_i` are cursors
* that iterate over the vertical and horizontal subsampling factors, respectively.
* When we finish one iteration of the innermost loop, we'll have the coefficients
* of one of the components of block at position `macroblock_index`. When the outermost
* loop finishes first iteration, we'll have all the luminance coefficients for all the
* macroblocks that share the chrominance data. Next two iterations (assuming that
* we are dealing with three components) will fill up the blocks with chroma data.
*/
static ErrorOr<void> build_macroblocks(JPEGLoadingContext& context, Vector<Macroblock>& macroblocks, u32 hcursor, u32 vcursor)
{
for (auto const& scan_component : context.current_scan.components) {
for (u8 vfactor_i = 0; vfactor_i < scan_component.component.vsample_factor; vfactor_i++) {
for (u8 hfactor_i = 0; hfactor_i < scan_component.component.hsample_factor; hfactor_i++) {
// A.2.3 - Interleaved order
u32 macroblock_index = (vcursor + vfactor_i) * context.mblock_meta.hpadded_count + (hfactor_i + hcursor);
if (!context.current_scan.are_components_interleaved()) {
macroblock_index = vcursor * context.mblock_meta.hpadded_count + (hfactor_i + (hcursor * scan_component.component.vsample_factor) + (vfactor_i * scan_component.component.hsample_factor));
// A.2.4 Completion of partial MCU
// If the component is [and only if!] to be interleaved, the encoding process
// shall also extend the number of samples by one or more additional blocks.
// Horizontally
if (macroblock_index >= context.mblock_meta.hcount && macroblock_index % context.mblock_meta.hpadded_count >= context.mblock_meta.hcount)
continue;
// Vertically
if (macroblock_index >= context.mblock_meta.hpadded_count * context.mblock_meta.vcount)
continue;
}
Macroblock& block = macroblocks[macroblock_index];
if (context.current_scan.spectral_selection_start == 0)
TRY(add_dc(context, block, scan_component));
if (context.current_scan.spectral_selection_end != 0)
TRY(add_ac(context, block, scan_component));
// G.1.2.2 - Progressive encoding of AC coefficients with Huffman coding
if (context.current_scan.end_of_bands_run_count > 0) {
--context.current_scan.end_of_bands_run_count;
continue;
}
}
}
}
return {};
}
static bool is_dct_based(StartOfFrame::FrameType frame_type)
{
return frame_type == StartOfFrame::FrameType::Baseline_DCT
|| frame_type == StartOfFrame::FrameType::Extended_Sequential_DCT
|| frame_type == StartOfFrame::FrameType::Progressive_DCT
|| frame_type == StartOfFrame::FrameType::Differential_Sequential_DCT
|| frame_type == StartOfFrame::FrameType::Differential_Progressive_DCT
|| frame_type == StartOfFrame::FrameType::Progressive_DCT_Arithmetic
|| frame_type == StartOfFrame::FrameType::Differential_Sequential_DCT_Arithmetic
|| frame_type == StartOfFrame::FrameType::Differential_Progressive_DCT_Arithmetic;
}
static void reset_decoder(JPEGLoadingContext& context)
{
// G.1.2.2 - Progressive encoding of AC coefficients with Huffman coding
context.current_scan.end_of_bands_run_count = 0;
// E.2.4 Control procedure for decoding a restart interval
if (is_dct_based(context.frame.type)) {
context.previous_dc_values = {};
return;
}
VERIFY_NOT_REACHED();
}
static ErrorOr<void> decode_huffman_stream(JPEGLoadingContext& context, Vector<Macroblock>& macroblocks)
{
for (u32 vcursor = 0; vcursor < context.mblock_meta.vcount; vcursor += context.vsample_factor) {
for (u32 hcursor = 0; hcursor < context.mblock_meta.hcount; hcursor += context.hsample_factor) {
u32 i = vcursor * context.mblock_meta.hpadded_count + hcursor;
auto& huffman_stream = context.current_scan.huffman_stream;
if (context.dc_restart_interval > 0) {
if (i != 0 && i % (context.dc_restart_interval * context.vsample_factor * context.hsample_factor) == 0) {
reset_decoder(context);
// Restart markers are stored in byte boundaries. Advance the huffman stream cursor to
// the 0th bit of the next byte.
huffman_stream.advance_to_byte_boundary();
// Skip the restart marker (RSTn).
huffman_stream.skip_byte();
}
}
if (auto result = build_macroblocks(context, macroblocks, hcursor, vcursor); result.is_error()) {
if constexpr (JPEG_DEBUG) {
dbgln("Failed to build Macroblock {}: {}", i, result.error());
dbgln("Huffman stream byte offset {}", huffman_stream.byte_offset());
}
return result.release_error();
}
}
}
return {};
}
static bool is_frame_marker(Marker const marker)
{
// B.1.1.3 - Marker assignments
bool const is_sof_marker = marker >= JPEG_SOF0 && marker <= JPEG_SOF15;
// Start of frame markers are valid for JPEG_SOF0 to JPEG_SOF15 except number 4, 8 (reserved) and 12.
bool const is_defined_marker = marker != JPEG_DHT && marker != 0xFFC8 && marker != JPEG_DAC;
return is_sof_marker && is_defined_marker;
}
static inline bool is_supported_marker(Marker const marker)
{
if (marker >= JPEG_APPN0 && marker <= JPEG_APPN15) {
if (marker != JPEG_APPN0 && marker != JPEG_APPN14)
dbgln_if(JPEG_DEBUG, "{:#04x} not supported yet. The decoder may fail!", marker);
return true;
}
if (marker >= JPEG_RESERVED1 && marker <= JPEG_RESERVEDD)
return true;
if (marker >= JPEG_RST0 && marker <= JPEG_RST7)
return true;
switch (marker) {
case JPEG_COM:
case JPEG_DHP:
case JPEG_EXP:
case JPEG_DHT:
case JPEG_DQT:
case JPEG_DRI:
case JPEG_EOI:
case JPEG_SOF0:
case JPEG_SOF1:
case JPEG_SOF2:
case JPEG_SOI:
case JPEG_SOS:
return true;
}
if (is_frame_marker(marker))
dbgln_if(JPEG_DEBUG, "Decoding this frame-type (SOF{}) is not currently supported. Decoder will fail!", marker & 0xf);
return false;
}
static inline ErrorOr<Marker> read_marker_at_cursor(Stream& stream)
{
u16 marker = TRY(stream.read_value<BigEndian<u16>>());
if (is_supported_marker(marker))
return marker;
if (marker != 0xFFFF)
return JPEG_INVALID;
u8 next;
do {
next = TRY(stream.read_value<u8>());
if (next == 0x00)
return JPEG_INVALID;
} while (next == 0xFF);
marker = 0xFF00 | (u16)next;
return is_supported_marker(marker) ? marker : JPEG_INVALID;
}
static ErrorOr<u16> read_effective_chunk_size(Stream& stream)
{
// The stored chunk size includes the size of `stored_size` itself.
u16 const stored_size = TRY(stream.read_value<BigEndian<u16>>());
if (stored_size < 2)
return Error::from_string_literal("Stored chunk size is too small");
return stored_size - 2;
}
static ErrorOr<void> read_start_of_scan(Stream& stream, JPEGLoadingContext& context)
{
// B.2.3 - Scan header syntax
if (context.state < JPEGLoadingContext::State::FrameDecoded)
return Error::from_string_literal("SOS found before reading a SOF");
[[maybe_unused]] u16 const bytes_to_read = TRY(read_effective_chunk_size(stream));
u8 const component_count = TRY(stream.read_value<u8>());
Scan current_scan;
Optional<u8> last_read;
u8 component_read = 0;
for (auto& component : context.components) {
// See the Csj paragraph:
// [...] the ordering in the scan header shall follow the ordering in the frame header.
if (component_read == component_count)
break;
if (!last_read.has_value())
last_read = TRY(stream.read_value<u8>());
if (component.id != *last_read)
continue;
u8 table_ids = TRY(stream.read_value<u8>());
current_scan.components.empend(component, static_cast<u8>(table_ids >> 4), static_cast<u8>(table_ids & 0x0F));
component_read++;
last_read.clear();
}
if constexpr (JPEG_DEBUG) {
StringBuilder builder;
TRY(builder.try_append("Components in scan: "sv));
for (auto const& scan_component : current_scan.components) {
TRY(builder.try_append(TRY(String::number(scan_component.component.id))));
TRY(builder.try_append(' '));
}
dbgln(builder.string_view());
}
current_scan.spectral_selection_start = TRY(stream.read_value<u8>());
current_scan.spectral_selection_end = TRY(stream.read_value<u8>());
auto const successive_approximation = TRY(stream.read_value<u8>());
current_scan.successive_approximation_high = successive_approximation >> 4;
current_scan.successive_approximation_low = successive_approximation & 0x0F;
dbgln_if(JPEG_DEBUG, "Start of Selection: {}, End of Selection: {}, Successive Approximation High: {}, Successive Approximation Low: {}",
current_scan.spectral_selection_start,
current_scan.spectral_selection_end,
current_scan.successive_approximation_high,
current_scan.successive_approximation_low);
if (current_scan.spectral_selection_start > 63 || current_scan.spectral_selection_end > 63 || current_scan.successive_approximation_high > 13 || current_scan.successive_approximation_low > 13) {
dbgln_if(JPEG_DEBUG, "ERROR! Start of Selection: {}, End of Selection: {}, Successive Approximation High: {}, Successive Approximation Low: {}!",
current_scan.spectral_selection_start,
current_scan.spectral_selection_end,
current_scan.successive_approximation_high,
current_scan.successive_approximation_low);
return Error::from_string_literal("Spectral selection is not [0,63] or successive approximation is not null");
}
current_scan.huffman_stream = TRY(HuffmanStream::create(*context.stream));
context.current_scan = move(current_scan);
return {};
}
static ErrorOr<void> read_restart_interval(Stream& stream, JPEGLoadingContext& context)
{
// B.2.4.4 - Restart interval definition syntax
u16 bytes_to_read = TRY(read_effective_chunk_size(stream));
if (bytes_to_read != 2) {
dbgln_if(JPEG_DEBUG, "Malformed DRI marker found!");
return Error::from_string_literal("Malformed DRI marker found");
}
context.dc_restart_interval = TRY(stream.read_value<BigEndian<u16>>());
return {};
}
static ErrorOr<void> read_huffman_table(Stream& stream, JPEGLoadingContext& context)
{
// B.2.4.2 - Huffman table-specification syntax
u16 bytes_to_read = TRY(read_effective_chunk_size(stream));
while (bytes_to_read > 0) {
HuffmanTable table;
u8 const table_info = TRY(stream.read_value<u8>());
u8 const table_type = table_info >> 4;
u8 const table_destination_id = table_info & 0x0F;
if (table_type > 1) {
dbgln_if(JPEG_DEBUG, "Unrecognized huffman table: {}!", table_type);
return Error::from_string_literal("Unrecognized huffman table");
}
if ((context.frame.type == StartOfFrame::FrameType::Baseline_DCT && table_destination_id > 1)
|| (context.frame.type != StartOfFrame::FrameType::Baseline_DCT && table_destination_id > 3)) {
dbgln_if(JPEG_DEBUG, "Invalid huffman table destination id: {}!", table_destination_id);
return Error::from_string_literal("Invalid huffman table destination id");
}
table.type = table_type;
table.destination_id = table_destination_id;
u32 total_codes = 0;
// Read code counts. At each index K, the value represents the number of K+1 bit codes in this header.
for (int i = 0; i < 16; i++) {
if (i == HuffmanTable::bits_per_cached_code)
table.first_non_cached_code_index = total_codes;
u8 count = TRY(stream.read_value<u8>());
total_codes += count;
table.code_counts[i] = count;
}
table.codes.ensure_capacity(total_codes);
table.symbols.ensure_capacity(total_codes);
// Read symbols. Read X bytes, where X is the sum of the counts of codes read in the previous step.
for (u32 i = 0; i < total_codes; i++) {
u8 symbol = TRY(stream.read_value<u8>());
table.symbols.append(symbol);
}
table.generate_codes();
auto& huffman_table = table.type == 0 ? context.dc_tables : context.ac_tables;
huffman_table.set(table.destination_id, table);
bytes_to_read -= 1 + 16 + total_codes;
}
if (bytes_to_read != 0) {
dbgln_if(JPEG_DEBUG, "Extra bytes detected in huffman header!");
return Error::from_string_literal("Extra bytes detected in huffman header");
}
return {};
}
static ErrorOr<void> read_icc_profile(Stream& stream, JPEGLoadingContext& context, int bytes_to_read)
{
// https://www.color.org/technotes/ICC-Technote-ProfileEmbedding.pdf, page 5, "JFIF".
if (bytes_to_read <= 2)
return Error::from_string_literal("icc marker too small");
auto chunk_sequence_number = TRY(stream.read_value<u8>()); // 1-based
auto number_of_chunks = TRY(stream.read_value<u8>());
bytes_to_read -= 2;
if (!context.icc_multi_chunk_state.has_value())
context.icc_multi_chunk_state.emplace(ICCMultiChunkState { 0, TRY(FixedArray<ByteBuffer>::create(number_of_chunks)) });
auto& chunk_state = context.icc_multi_chunk_state;
if (chunk_state->seen_number_of_icc_chunks >= number_of_chunks)
return Error::from_string_literal("Too many ICC chunks");
if (chunk_state->chunks.size() != number_of_chunks)
return Error::from_string_literal("Inconsistent number of total ICC chunks");
if (chunk_sequence_number == 0)
return Error::from_string_literal("ICC chunk sequence number not 1 based");
u8 index = chunk_sequence_number - 1;
if (index >= chunk_state->chunks.size())
return Error::from_string_literal("ICC chunk sequence number larger than number of chunks");
if (!chunk_state->chunks[index].is_empty())
return Error::from_string_literal("Duplicate ICC chunk at sequence number");
chunk_state->chunks[index] = TRY(ByteBuffer::create_zeroed(bytes_to_read));
TRY(stream.read_until_filled(chunk_state->chunks[index]));
chunk_state->seen_number_of_icc_chunks++;
if (chunk_state->seen_number_of_icc_chunks != chunk_state->chunks.size())
return {};
if (number_of_chunks == 1) {
context.icc_data = move(chunk_state->chunks[0]);
return {};
}
size_t total_size = 0;
for (auto const& chunk : chunk_state->chunks)
total_size += chunk.size();
auto icc_bytes = TRY(ByteBuffer::create_zeroed(total_size));
size_t start = 0;
for (auto const& chunk : chunk_state->chunks) {
memcpy(icc_bytes.data() + start, chunk.data(), chunk.size());
start += chunk.size();
}
context.icc_data = move(icc_bytes);
return {};
}
static ErrorOr<void> read_colour_encoding(Stream& stream, [[maybe_unused]] JPEGLoadingContext& context, int bytes_to_read)
{
// The App 14 segment is application specific in the first JPEG standard.
// However, the Adobe implementation is globally accepted and the value of the color transform
// was latter standardized as a JPEG-1 extension.
// For the structure of the App 14 segment, see:
// https://www.pdfa.org/norm-refs/5116.DCT_Filter.pdf
// 18 Adobe Application-Specific JPEG Marker
// For the value of color_transform, see:
// https://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-T.872-201206-I!!PDF-E&type=items
// 6.5.3 - APP14 marker segment for colour encoding
if (bytes_to_read < 6)
return Error::from_string_literal("App14 segment too small");
[[maybe_unused]] auto const version = TRY(stream.read_value<u8>());
[[maybe_unused]] u16 const flag0 = TRY(stream.read_value<BigEndian<u16>>());
[[maybe_unused]] u16 const flag1 = TRY(stream.read_value<BigEndian<u16>>());
auto const color_transform = TRY(stream.read_value<u8>());
if (bytes_to_read > 6) {
dbgln_if(JPEG_DEBUG, "Unread bytes in App14 segment: {}", bytes_to_read - 6);
TRY(stream.discard(bytes_to_read - 6));
}
switch (color_transform) {
case 0:
context.color_transform = ColorTransform::CmykOrRgb;
break;
case 1:
context.color_transform = ColorTransform::YCbCr;
break;
case 2:
context.color_transform = ColorTransform::YCCK;
break;
default:
dbgln("0x{:x} is not a specified transform flag value, ignoring", color_transform);
}
return {};
}
static ErrorOr<void> read_app_marker(Stream& stream, JPEGLoadingContext& context, int app_marker_number)
{
// B.2.4.6 - Application data syntax
u16 bytes_to_read = TRY(read_effective_chunk_size(stream));
StringBuilder builder;
for (;;) {
if (bytes_to_read == 0) {
dbgln_if(JPEG_DEBUG, "app marker {} does not start with zero-terminated string", app_marker_number);
return {};
}
auto c = TRY(stream.read_value<char>());
bytes_to_read--;
if (c == '\0')
break;
TRY(builder.try_append(c));
}
auto app_id = TRY(builder.to_string());
if (app_marker_number == 2 && app_id == "ICC_PROFILE"sv)
return read_icc_profile(stream, context, bytes_to_read);
if (app_marker_number == 14 && app_id == "Adobe"sv)
return read_colour_encoding(stream, context, bytes_to_read);
return stream.discard(bytes_to_read);
}
static inline bool validate_luma_and_modify_context(Component const& luma, JPEGLoadingContext& context)
{
if ((luma.hsample_factor == 1 || luma.hsample_factor == 2) && (luma.vsample_factor == 1 || luma.vsample_factor == 2)) {
context.mblock_meta.hpadded_count += luma.hsample_factor == 1 ? 0 : context.mblock_meta.hcount % 2;
context.mblock_meta.vpadded_count += luma.vsample_factor == 1 ? 0 : context.mblock_meta.vcount % 2;
context.mblock_meta.padded_total = context.mblock_meta.hpadded_count * context.mblock_meta.vpadded_count;
// For easy reference to relevant sample factors.
context.hsample_factor = luma.hsample_factor;
context.vsample_factor = luma.vsample_factor;
if constexpr (JPEG_DEBUG) {
dbgln("Horizontal Subsampling Factor: {}", luma.hsample_factor);
dbgln("Vertical Subsampling Factor: {}", luma.vsample_factor);
}
return true;
}
return false;
}
static inline void set_macroblock_metadata(JPEGLoadingContext& context)
{
context.mblock_meta.hcount = (context.frame.width + 7) / 8;
context.mblock_meta.vcount = (context.frame.height + 7) / 8;
context.mblock_meta.hpadded_count = context.mblock_meta.hcount;
context.mblock_meta.vpadded_count = context.mblock_meta.vcount;
context.mblock_meta.total = context.mblock_meta.hcount * context.mblock_meta.vcount;
}
static ErrorOr<void> ensure_standard_precision(StartOfFrame const& frame)
{
// B.2.2 - Frame header syntax
// Table B.2 - Frame header parameter sizes and values
if (frame.precision == 8)
return {};
if (frame.type == StartOfFrame::FrameType::Extended_Sequential_DCT && frame.precision == 12)
return {};
if (frame.type == StartOfFrame::FrameType::Progressive_DCT && frame.precision == 12)
return {};
dbgln_if(JPEG_DEBUG, "Unsupported precision: {}, for SOF type: {}!", frame.precision, static_cast<int>(frame.type));
return Error::from_string_literal("Unsupported SOF precision.");
}
static ErrorOr<void> read_start_of_frame(Stream& stream, JPEGLoadingContext& context)
{
if (context.state == JPEGLoadingContext::FrameDecoded) {
dbgln_if(JPEG_DEBUG, "SOF repeated!");
return Error::from_string_literal("SOF repeated");
}
[[maybe_unused]] u16 const bytes_to_read = TRY(read_effective_chunk_size(stream));
context.frame.precision = TRY(stream.read_value<u8>());
TRY(ensure_standard_precision(context.frame));
context.frame.height = TRY(stream.read_value<BigEndian<u16>>());
context.frame.width = TRY(stream.read_value<BigEndian<u16>>());
if (!context.frame.width || !context.frame.height) {
dbgln_if(JPEG_DEBUG, "ERROR! Image height: {}, Image width: {}!", context.frame.height, context.frame.width);
return Error::from_string_literal("Image frame height of width null");
}
if (context.frame.width > maximum_width_for_decoded_images || context.frame.height > maximum_height_for_decoded_images) {
dbgln("This JPEG is too large for comfort: {}x{}", context.frame.width, context.frame.height);
return Error::from_string_literal("JPEG too large for comfort");
}
set_macroblock_metadata(context);
auto component_count = TRY(stream.read_value<u8>());
if (component_count != 1 && component_count != 3 && component_count != 4) {
dbgln_if(JPEG_DEBUG, "Unsupported number of components in SOF: {}!", component_count);
return Error::from_string_literal("Unsupported number of components in SOF");
}
for (u8 i = 0; i < component_count; i++) {
Component component;
component.id = TRY(stream.read_value<u8>());
component.index = i;
u8 subsample_factors = TRY(stream.read_value<u8>());
component.hsample_factor = subsample_factors >> 4;
component.vsample_factor = subsample_factors & 0x0F;
if (i == 0) {
// By convention, downsampling is applied only on chroma components. So we should
// hope to see the maximum sampling factor in the luma component.
if (!validate_luma_and_modify_context(component, context)) {
dbgln_if(JPEG_DEBUG, "Unsupported luma subsampling factors: horizontal: {}, vertical: {}",
component.hsample_factor,
component.vsample_factor);
return Error::from_string_literal("Unsupported luma subsampling factors");
}
} else {
if (component.hsample_factor != 1 || component.vsample_factor != 1) {
dbgln_if(JPEG_DEBUG, "Unsupported chroma subsampling factors: horizontal: {}, vertical: {}",
component.hsample_factor,
component.vsample_factor);
return Error::from_string_literal("Unsupported chroma subsampling factors");
}
}
component.quantization_table_id = TRY(stream.read_value<u8>());
context.components.append(move(component));
}
return {};
}
static ErrorOr<void> read_quantization_table(Stream& stream, JPEGLoadingContext& context)
{
// B.2.4.1 - Quantization table-specification syntax
u16 bytes_to_read = TRY(read_effective_chunk_size(stream));
while (bytes_to_read > 0) {
u8 const info_byte = TRY(stream.read_value<u8>());
u8 const element_unit_hint = info_byte >> 4;
if (element_unit_hint > 1) {
dbgln_if(JPEG_DEBUG, "Unsupported unit hint in quantization table: {}!", element_unit_hint);
return Error::from_string_literal("Unsupported unit hint in quantization table");
}
u8 const table_id = info_byte & 0x0F;
if (table_id > 3) {
dbgln_if(JPEG_DEBUG, "Unsupported quantization table id: {}!", table_id);
return Error::from_string_literal("Unsupported quantization table id");
}
auto& maybe_table = context.quantization_tables[table_id];
if (!maybe_table.has_value())
maybe_table = Array<u16, 64> {};
auto& table = maybe_table.value();
for (int i = 0; i < 64; i++) {
if (element_unit_hint == 0)
table[zigzag_map[i]] = TRY(stream.read_value<u8>());
else
table[zigzag_map[i]] = TRY(stream.read_value<BigEndian<u16>>());
}
bytes_to_read -= 1 + (element_unit_hint == 0 ? 64 : 128);
}
if (bytes_to_read != 0) {
dbgln_if(JPEG_DEBUG, "Invalid length for one or more quantization tables!");
return Error::from_string_literal("Invalid length for one or more quantization tables");
}
return {};
}
static ErrorOr<void> skip_segment(Stream& stream)
{
u16 bytes_to_skip = TRY(stream.read_value<BigEndian<u16>>()) - 2;
TRY(stream.discard(bytes_to_skip));
return {};
}
static ErrorOr<void> dequantize(JPEGLoadingContext& context, Vector<Macroblock>& macroblocks)
{
for (u32 vcursor = 0; vcursor < context.mblock_meta.vcount; vcursor += context.vsample_factor) {
for (u32 hcursor = 0; hcursor < context.mblock_meta.hcount; hcursor += context.hsample_factor) {
for (u32 i = 0; i < context.components.size(); i++) {
auto const& component = context.components[i];
if (!context.quantization_tables[component.quantization_table_id].has_value()) {
dbgln_if(JPEG_DEBUG, "Unknown quantization table id: {}!", component.quantization_table_id);
return Error::from_string_literal("Unknown quantization table id");
}
auto const& table = context.quantization_tables[component.quantization_table_id].value();
for (u32 vfactor_i = 0; vfactor_i < component.vsample_factor; vfactor_i++) {
for (u32 hfactor_i = 0; hfactor_i < component.hsample_factor; hfactor_i++) {
u32 macroblock_index = (vcursor + vfactor_i) * context.mblock_meta.hpadded_count + (hfactor_i + hcursor);
Macroblock& block = macroblocks[macroblock_index];
auto* block_component = get_component(block, i);
for (u32 k = 0; k < 64; k++)
block_component[k] *= table[k];
}
}
}
}
}
return {};
}
static void inverse_dct(JPEGLoadingContext const& context, Vector<Macroblock>& macroblocks)
{
static float const m0 = 2.0f * AK::cos(1.0f / 16.0f * 2.0f * AK::Pi<float>);
static float const m1 = 2.0f * AK::cos(2.0f / 16.0f * 2.0f * AK::Pi<float>);
static float const m3 = 2.0f * AK::cos(2.0f / 16.0f * 2.0f * AK::Pi<float>);
static float const m5 = 2.0f * AK::cos(3.0f / 16.0f * 2.0f * AK::Pi<float>);
static float const m2 = m0 - m5;
static float const m4 = m0 + m5;
static float const s0 = AK::cos(0.0f / 16.0f * AK::Pi<float>) * AK::rsqrt(8.0f);
static float const s1 = AK::cos(1.0f / 16.0f * AK::Pi<float>) / 2.0f;
static float const s2 = AK::cos(2.0f / 16.0f * AK::Pi<float>) / 2.0f;
static float const s3 = AK::cos(3.0f / 16.0f * AK::Pi<float>) / 2.0f;
static float const s4 = AK::cos(4.0f / 16.0f * AK::Pi<float>) / 2.0f;
static float const s5 = AK::cos(5.0f / 16.0f * AK::Pi<float>) / 2.0f;
static float const s6 = AK::cos(6.0f / 16.0f * AK::Pi<float>) / 2.0f;
static float const s7 = AK::cos(7.0f / 16.0f * AK::Pi<float>) / 2.0f;
for (u32 vcursor = 0; vcursor < context.mblock_meta.vcount; vcursor += context.vsample_factor) {
for (u32 hcursor = 0; hcursor < context.mblock_meta.hcount; hcursor += context.hsample_factor) {
for (u32 component_i = 0; component_i < context.components.size(); component_i++) {
auto& component = context.components[component_i];
for (u8 vfactor_i = 0; vfactor_i < component.vsample_factor; vfactor_i++) {
for (u8 hfactor_i = 0; hfactor_i < component.hsample_factor; hfactor_i++) {
u32 macroblock_index = (vcursor + vfactor_i) * context.mblock_meta.hpadded_count + (hfactor_i + hcursor);
Macroblock& block = macroblocks[macroblock_index];
auto* block_component = get_component(block, component_i);
for (u32 k = 0; k < 8; ++k) {
float const g0 = block_component[0 * 8 + k] * s0;
float const g1 = block_component[4 * 8 + k] * s4;
float const g2 = block_component[2 * 8 + k] * s2;
float const g3 = block_component[6 * 8 + k] * s6;
float const g4 = block_component[5 * 8 + k] * s5;
float const g5 = block_component[1 * 8 + k] * s1;
float const g6 = block_component[7 * 8 + k] * s7;
float const g7 = block_component[3 * 8 + k] * s3;
float const f0 = g0;
float const f1 = g1;
float const f2 = g2;
float const f3 = g3;
float const f4 = g4 - g7;
float const f5 = g5 + g6;
float const f6 = g5 - g6;
float const f7 = g4 + g7;
float const e0 = f0;
float const e1 = f1;
float const e2 = f2 - f3;
float const e3 = f2 + f3;
float const e4 = f4;
float const e5 = f5 - f7;
float const e6 = f6;
float const e7 = f5 + f7;
float const e8 = f4 + f6;
float const d0 = e0;
float const d1 = e1;
float const d2 = e2 * m1;
float const d3 = e3;
float const d4 = e4 * m2;
float const d5 = e5 * m3;
float const d6 = e6 * m4;
float const d7 = e7;
float const d8 = e8 * m5;
float const c0 = d0 + d1;
float const c1 = d0 - d1;
float const c2 = d2 - d3;
float const c3 = d3;
float const c4 = d4 + d8;
float const c5 = d5 + d7;
float const c6 = d6 - d8;
float const c7 = d7;
float const c8 = c5 - c6;
float const b0 = c0 + c3;
float const b1 = c1 + c2;
float const b2 = c1 - c2;
float const b3 = c0 - c3;
float const b4 = c4 - c8;
float const b5 = c8;
float const b6 = c6 - c7;
float const b7 = c7;
block_component[0 * 8 + k] = b0 + b7;
block_component[1 * 8 + k] = b1 + b6;
block_component[2 * 8 + k] = b2 + b5;
block_component[3 * 8 + k] = b3 + b4;
block_component[4 * 8 + k] = b3 - b4;
block_component[5 * 8 + k] = b2 - b5;
block_component[6 * 8 + k] = b1 - b6;
block_component[7 * 8 + k] = b0 - b7;
}
for (u32 l = 0; l < 8; ++l) {
float const g0 = block_component[l * 8 + 0] * s0;
float const g1 = block_component[l * 8 + 4] * s4;
float const g2 = block_component[l * 8 + 2] * s2;
float const g3 = block_component[l * 8 + 6] * s6;
float const g4 = block_component[l * 8 + 5] * s5;
float const g5 = block_component[l * 8 + 1] * s1;
float const g6 = block_component[l * 8 + 7] * s7;
float const g7 = block_component[l * 8 + 3] * s3;
float const f0 = g0;
float const f1 = g1;
float const f2 = g2;
float const f3 = g3;
float const f4 = g4 - g7;
float const f5 = g5 + g6;
float const f6 = g5 - g6;
float const f7 = g4 + g7;
float const e0 = f0;
float const e1 = f1;
float const e2 = f2 - f3;
float const e3 = f2 + f3;
float const e4 = f4;
float const e5 = f5 - f7;
float const e6 = f6;
float const e7 = f5 + f7;
float const e8 = f4 + f6;
float const d0 = e0;
float const d1 = e1;
float const d2 = e2 * m1;
float const d3 = e3;
float const d4 = e4 * m2;
float const d5 = e5 * m3;
float const d6 = e6 * m4;
float const d7 = e7;
float const d8 = e8 * m5;
float const c0 = d0 + d1;
float const c1 = d0 - d1;
float const c2 = d2 - d3;
float const c3 = d3;
float const c4 = d4 + d8;
float const c5 = d5 + d7;
float const c6 = d6 - d8;
float const c7 = d7;
float const c8 = c5 - c6;
float const b0 = c0 + c3;
float const b1 = c1 + c2;
float const b2 = c1 - c2;
float const b3 = c0 - c3;
float const b4 = c4 - c8;
float const b5 = c8;
float const b6 = c6 - c7;
float const b7 = c7;
block_component[l * 8 + 0] = b0 + b7;
block_component[l * 8 + 1] = b1 + b6;
block_component[l * 8 + 2] = b2 + b5;
block_component[l * 8 + 3] = b3 + b4;
block_component[l * 8 + 4] = b3 - b4;
block_component[l * 8 + 5] = b2 - b5;
block_component[l * 8 + 6] = b1 - b6;
block_component[l * 8 + 7] = b0 - b7;
}
}
}
}
}
}
// F.2.1.5 - Inverse DCT (IDCT)
auto const level_shift = 1 << (context.frame.precision - 1);
auto const max_value = (1 << context.frame.precision) - 1;
for (u32 vcursor = 0; vcursor < context.mblock_meta.vcount; vcursor += context.vsample_factor) {
for (u32 hcursor = 0; hcursor < context.mblock_meta.hcount; hcursor += context.hsample_factor) {
for (u8 vfactor_i = 0; vfactor_i < context.vsample_factor; ++vfactor_i) {
for (u8 hfactor_i = 0; hfactor_i < context.hsample_factor; ++hfactor_i) {
u32 mb_index = (vcursor + vfactor_i) * context.mblock_meta.hpadded_count + (hcursor + hfactor_i);
for (u8 i = 0; i < 8; ++i) {
for (u8 j = 0; j < 8; ++j) {
// FIXME: This just truncate all coefficients, it's an easy way to support (read hack)
// 12 bits JPEGs without rewriting all color transformations.
auto const clamp_to_8_bits = [&](u16 color) -> u8 {
if (context.frame.precision == 8)
return static_cast<u8>(color);
return static_cast<u8>(color >> 4);
};
macroblocks[mb_index].r[i * 8 + j] = clamp_to_8_bits(clamp(macroblocks[mb_index].r[i * 8 + j] + level_shift, 0, max_value));
macroblocks[mb_index].g[i * 8 + j] = clamp_to_8_bits(clamp(macroblocks[mb_index].g[i * 8 + j] + level_shift, 0, max_value));
macroblocks[mb_index].b[i * 8 + j] = clamp_to_8_bits(clamp(macroblocks[mb_index].b[i * 8 + j] + level_shift, 0, max_value));
macroblocks[mb_index].k[i * 8 + j] = clamp_to_8_bits(clamp(macroblocks[mb_index].k[i * 8 + j] + level_shift, 0, max_value));
}
}
}
}
}
}
}
static void ycbcr_to_rgb(JPEGLoadingContext const& context, Vector<Macroblock>& macroblocks)
{
// Conversion from YCbCr to RGB isn't specified in the first JPEG specification but in the JFIF extension:
// See: https://www.itu.int/rec/dologin_pub.asp?lang=f&id=T-REC-T.871-201105-I!!PDF-E&type=items
// 7 - Conversion to and from RGB
for (u32 vcursor = 0; vcursor < context.mblock_meta.vcount; vcursor += context.vsample_factor) {
for (u32 hcursor = 0; hcursor < context.mblock_meta.hcount; hcursor += context.hsample_factor) {
const u32 chroma_block_index = vcursor * context.mblock_meta.hpadded_count + hcursor;
Macroblock const& chroma = macroblocks[chroma_block_index];
// Overflows are intentional.
for (u8 vfactor_i = context.vsample_factor - 1; vfactor_i < context.vsample_factor; --vfactor_i) {
for (u8 hfactor_i = context.hsample_factor - 1; hfactor_i < context.hsample_factor; --hfactor_i) {
u32 macroblock_index = (vcursor + vfactor_i) * context.mblock_meta.hpadded_count + (hcursor + hfactor_i);
auto* y = macroblocks[macroblock_index].y;
auto* cb = macroblocks[macroblock_index].cb;
auto* cr = macroblocks[macroblock_index].cr;
for (u8 i = 7; i < 8; --i) {
for (u8 j = 7; j < 8; --j) {
const u8 pixel = i * 8 + j;
const u32 chroma_pxrow = (i / context.vsample_factor) + 4 * vfactor_i;
const u32 chroma_pxcol = (j / context.hsample_factor) + 4 * hfactor_i;
const u32 chroma_pixel = chroma_pxrow * 8 + chroma_pxcol;
int r = y[pixel] + 1.402f * (chroma.cr[chroma_pixel] - 128);
int g = y[pixel] - 0.3441f * (chroma.cb[chroma_pixel] - 128) - 0.7141f * (chroma.cr[chroma_pixel] - 128);
int b = y[pixel] + 1.772f * (chroma.cb[chroma_pixel] - 128);
y[pixel] = clamp(r, 0, 255);
cb[pixel] = clamp(g, 0, 255);
cr[pixel] = clamp(b, 0, 255);
}
}
}
}
}
}
}
static void invert_colors_for_adobe_images(JPEGLoadingContext const& context, Vector<Macroblock>& macroblocks)
{
if (!context.color_transform.has_value())
return;
// From libjpeg-turbo's libjpeg.txt:
// https://github.com/libjpeg-turbo/libjpeg-turbo/blob/main/libjpeg.txt
// CAUTION: it appears that Adobe Photoshop writes inverted data in CMYK JPEG
// files: 0 represents 100% ink coverage, rather than 0% ink as you'd expect.
// This is arguably a bug in Photoshop, but if you need to work with Photoshop
// CMYK files, you will have to deal with it in your application.
for (u32 vcursor = 0; vcursor < context.mblock_meta.vcount; vcursor += context.vsample_factor) {
for (u32 hcursor = 0; hcursor < context.mblock_meta.hcount; hcursor += context.hsample_factor) {
for (u8 vfactor_i = 0; vfactor_i < context.vsample_factor; ++vfactor_i) {
for (u8 hfactor_i = 0; hfactor_i < context.hsample_factor; ++hfactor_i) {
u32 mb_index = (vcursor + vfactor_i) * context.mblock_meta.hpadded_count + (hcursor + hfactor_i);
for (u8 i = 0; i < 8; ++i) {
for (u8 j = 0; j < 8; ++j) {
macroblocks[mb_index].r[i * 8 + j] = NumericLimits<u8>::max() - macroblocks[mb_index].r[i * 8 + j];
macroblocks[mb_index].g[i * 8 + j] = NumericLimits<u8>::max() - macroblocks[mb_index].g[i * 8 + j];
macroblocks[mb_index].b[i * 8 + j] = NumericLimits<u8>::max() - macroblocks[mb_index].b[i * 8 + j];
macroblocks[mb_index].k[i * 8 + j] = NumericLimits<u8>::max() - macroblocks[mb_index].k[i * 8 + j];
}
}
}
}
}
}
}
static void cmyk_to_rgb(JPEGLoadingContext const& context, Vector<Macroblock>& macroblocks)
{
invert_colors_for_adobe_images(context, macroblocks);
for (u32 vcursor = 0; vcursor < context.mblock_meta.vcount; vcursor += context.vsample_factor) {
for (u32 hcursor = 0; hcursor < context.mblock_meta.hcount; hcursor += context.hsample_factor) {
for (u8 vfactor_i = context.vsample_factor - 1; vfactor_i < context.vsample_factor; --vfactor_i) {
for (u8 hfactor_i = context.hsample_factor - 1; hfactor_i < context.hsample_factor; --hfactor_i) {
u32 mb_index = (vcursor + vfactor_i) * context.mblock_meta.hpadded_count + (hcursor + hfactor_i);
auto* c = macroblocks[mb_index].y;
auto* m = macroblocks[mb_index].cb;
auto* y = macroblocks[mb_index].cr;
auto* k = macroblocks[mb_index].k;
for (u8 i = 0; i < 8; ++i) {
for (u8 j = 0; j < 8; ++j) {
u8 const pixel = i * 8 + j;
static constexpr auto max_value = NumericLimits<u8>::max();
auto const black_component = max_value - k[pixel];
int const r = ((max_value - c[pixel]) * black_component) / max_value;
int const g = ((max_value - m[pixel]) * black_component) / max_value;
int const b = ((max_value - y[pixel]) * black_component) / max_value;
c[pixel] = clamp(r, 0, max_value);
m[pixel] = clamp(g, 0, max_value);
y[pixel] = clamp(b, 0, max_value);
}
}
}
}
}
}
}
static void ycck_to_rgb(JPEGLoadingContext const& context, Vector<Macroblock>& macroblocks)
{
// 7 - Conversions between colour encodings
// YCCK is obtained from CMYK by converting the CMY channels to YCC channel.
// To convert back into RGB, we only need the 3 first components, which are baseline YCbCr
ycbcr_to_rgb(context, macroblocks);
// RGB to CMYK, as mentioned in https://www.smcm.iqfr.csic.es/docs/intel/ipp/ipp_manual/IPPI/ippi_ch15/functn_YCCKToCMYK_JPEG.htm#functn_YCCKToCMYK_JPEG
for (u32 vcursor = 0; vcursor < context.mblock_meta.vcount; vcursor += context.vsample_factor) {
for (u32 hcursor = 0; hcursor < context.mblock_meta.hcount; hcursor += context.hsample_factor) {
for (u8 vfactor_i = 0; vfactor_i < context.vsample_factor; ++vfactor_i) {
for (u8 hfactor_i = 0; hfactor_i < context.hsample_factor; ++hfactor_i) {
u32 mb_index = (vcursor + vfactor_i) * context.mblock_meta.hpadded_count + (hcursor + hfactor_i);
for (u8 i = 0; i < 8; ++i) {
for (u8 j = 0; j < 8; ++j) {
macroblocks[mb_index].r[i * 8 + j] = NumericLimits<u8>::max() - macroblocks[mb_index].r[i * 8 + j];
macroblocks[mb_index].g[i * 8 + j] = NumericLimits<u8>::max() - macroblocks[mb_index].g[i * 8 + j];
macroblocks[mb_index].b[i * 8 + j] = NumericLimits<u8>::max() - macroblocks[mb_index].b[i * 8 + j];
}
}
}
}
}
}
cmyk_to_rgb(context, macroblocks);
}
static ErrorOr<void> handle_color_transform(JPEGLoadingContext const& context, Vector<Macroblock>& macroblocks)
{
if (context.color_transform.has_value()) {
// https://www.itu.int/rec/dologin_pub.asp?lang=e&id=T-REC-T.872-201206-I!!PDF-E&type=items
// 6.5.3 - APP14 marker segment for colour encoding
switch (*context.color_transform) {
case ColorTransform::CmykOrRgb:
if (context.components.size() == 4) {
cmyk_to_rgb(context, macroblocks);
} else if (context.components.size() == 3) {
// Note: components.size() == 3 means that we have an RGB image, so no color transformation is needed.
} else {
return Error::from_string_literal("Wrong number of components for CMYK or RGB, aborting.");
}
break;
case ColorTransform::YCbCr:
ycbcr_to_rgb(context, macroblocks);
break;
case ColorTransform::YCCK:
ycck_to_rgb(context, macroblocks);
break;
}
return {};
}
// No App14 segment is present, assuming :
// - 1 components means grayscale
// - 3 components means YCbCr
// - 4 components means CMYK
if (context.components.size() == 4)
cmyk_to_rgb(context, macroblocks);
if (context.components.size() == 3)
ycbcr_to_rgb(context, macroblocks);
if (context.components.size() == 1) {
// With Cb and Cr being equal to zero, this function assign the Y
// value (luminosity) to R, G and B. Providing a proper conversion
// from grayscale to RGB.
ycbcr_to_rgb(context, macroblocks);
}
return {};
}
static ErrorOr<void> compose_bitmap(JPEGLoadingContext& context, Vector<Macroblock> const& macroblocks)
{
context.bitmap = TRY(Bitmap::create(BitmapFormat::BGRx8888, { context.frame.width, context.frame.height }));
for (u32 y = context.frame.height - 1; y < context.frame.height; y--) {
const u32 block_row = y / 8;
const u32 pixel_row = y % 8;
for (u32 x = 0; x < context.frame.width; x++) {
const u32 block_column = x / 8;
auto& block = macroblocks[block_row * context.mblock_meta.hpadded_count + block_column];
const u32 pixel_column = x % 8;
const u32 pixel_index = pixel_row * 8 + pixel_column;
const Color color { (u8)block.y[pixel_index], (u8)block.cb[pixel_index], (u8)block.cr[pixel_index] };
context.bitmap->set_pixel(x, y, color);
}
}
return {};
}
static bool is_app_marker(Marker const marker)
{
return marker >= JPEG_APPN0 && marker <= JPEG_APPN15;
}
static bool is_miscellaneous_or_table_marker(Marker const marker)
{
// B.2.4 - Table-specification and miscellaneous marker segment syntax
// See also B.6 - Summary: Figure B.17 – Flow of marker segment
bool const is_misc = marker == JPEG_COM || marker == JPEG_DRI || is_app_marker(marker);
bool const is_table = marker == JPEG_DQT || marker == JPEG_DAC || marker == JPEG_DHT;
return is_misc || is_table;
}
static ErrorOr<void> handle_miscellaneous_or_table(Stream& stream, JPEGLoadingContext& context, Marker const marker)
{
if (is_app_marker(marker)) {
TRY(read_app_marker(stream, context, marker - JPEG_APPN0));
return {};
}
switch (marker) {
case JPEG_COM:
case JPEG_DAC:
dbgln_if(JPEG_DEBUG, "TODO: implement marker \"{:x}\"", marker);
if (auto result = skip_segment(stream); result.is_error()) {
dbgln_if(JPEG_DEBUG, "Error skipping marker: {:x}!", marker);
return result.release_error();
}
break;
case JPEG_DHT:
TRY(read_huffman_table(stream, context));
break;
case JPEG_DQT:
TRY(read_quantization_table(stream, context));
break;
case JPEG_DRI:
TRY(read_restart_interval(stream, context));
break;
default:
dbgln("Unexpected marker: {:x}", marker);
VERIFY_NOT_REACHED();
}
return {};
}
static ErrorOr<void> parse_header(Stream& stream, JPEGLoadingContext& context)
{
auto marker = TRY(read_marker_at_cursor(stream));
if (marker != JPEG_SOI) {
dbgln_if(JPEG_DEBUG, "SOI not found: {:x}!", marker);
return Error::from_string_literal("SOI not found");
}
for (;;) {
marker = TRY(read_marker_at_cursor(stream));
if (is_miscellaneous_or_table_marker(marker)) {
TRY(handle_miscellaneous_or_table(stream, context, marker));
continue;
}
// Set frame type if the marker marks a new frame.
if (is_frame_marker(marker))
context.frame.type = static_cast<StartOfFrame::FrameType>(marker & 0xF);
switch (marker) {
case JPEG_INVALID:
case JPEG_RST0:
case JPEG_RST1:
case JPEG_RST2:
case JPEG_RST3:
case JPEG_RST4:
case JPEG_RST5:
case JPEG_RST6:
case JPEG_RST7:
case JPEG_SOI:
case JPEG_EOI:
dbgln_if(JPEG_DEBUG, "Unexpected marker {:x}!", marker);
return Error::from_string_literal("Unexpected marker");
case JPEG_SOF0:
case JPEG_SOF1:
case JPEG_SOF2:
TRY(read_start_of_frame(stream, context));
context.state = JPEGLoadingContext::FrameDecoded;
return {};
default:
if (auto result = skip_segment(stream); result.is_error()) {
dbgln_if(JPEG_DEBUG, "Error skipping marker: {:x}!", marker);
return result.release_error();
}
break;
}
}
VERIFY_NOT_REACHED();
}
static ErrorOr<void> decode_header(JPEGLoadingContext& context)
{
if (context.state < JPEGLoadingContext::State::HeaderDecoded) {
if (auto result = parse_header(*context.stream, context); result.is_error()) {
context.state = JPEGLoadingContext::State::Error;
return result.release_error();
}
if constexpr (JPEG_DEBUG) {
dbgln("Image width: {}", context.frame.width);
dbgln("Image height: {}", context.frame.height);
dbgln("Macroblocks in a row: {}", context.mblock_meta.hpadded_count);
dbgln("Macroblocks in a column: {}", context.mblock_meta.vpadded_count);
dbgln("Macroblock meta padded total: {}", context.mblock_meta.padded_total);
}
context.state = JPEGLoadingContext::State::HeaderDecoded;
}
return {};
}
static ErrorOr<Vector<Macroblock>> construct_macroblocks(JPEGLoadingContext& context)
{
// B.6 - Summary
// See: Figure B.16 – Flow of compressed data syntax
// This function handles the "Multi-scan" loop.
Vector<Macroblock> macroblocks;
TRY(macroblocks.try_resize(context.mblock_meta.padded_total));
Marker marker = TRY(read_marker_at_cursor(*context.stream));
while (true) {
if (is_miscellaneous_or_table_marker(marker)) {
TRY(handle_miscellaneous_or_table(*context.stream, context, marker));
} else if (marker == JPEG_SOS) {
TRY(read_start_of_scan(*context.stream, context));
TRY(decode_huffman_stream(context, macroblocks));
} else if (marker == JPEG_EOI) {
return macroblocks;
} else {
dbgln_if(JPEG_DEBUG, "Unexpected marker {:x}!", marker);
return Error::from_string_literal("Unexpected marker");
}
marker = TRY(read_marker_at_cursor(*context.stream));
}
}
static ErrorOr<void> decode_jpeg(JPEGLoadingContext& context)
{
TRY(decode_header(context));
auto macroblocks = TRY(construct_macroblocks(context));
TRY(dequantize(context, macroblocks));
inverse_dct(context, macroblocks);
TRY(handle_color_transform(context, macroblocks));
TRY(compose_bitmap(context, macroblocks));
context.stream.clear();
return {};
}
JPEGImageDecoderPlugin::JPEGImageDecoderPlugin(NonnullOwnPtr<FixedMemoryStream> stream)
{
m_context = make<JPEGLoadingContext>();
m_context->stream = move(stream);
}
JPEGImageDecoderPlugin::~JPEGImageDecoderPlugin() = default;
IntSize JPEGImageDecoderPlugin::size()
{
if (m_context->state == JPEGLoadingContext::State::Error)
return {};
if (m_context->state >= JPEGLoadingContext::State::FrameDecoded)
return { m_context->frame.width, m_context->frame.height };
return {};
}
void JPEGImageDecoderPlugin::set_volatile()
{
if (m_context->bitmap)
m_context->bitmap->set_volatile();
}
bool JPEGImageDecoderPlugin::set_nonvolatile(bool& was_purged)
{
if (!m_context->bitmap)
return false;
return m_context->bitmap->set_nonvolatile(was_purged);
}
ErrorOr<void> JPEGImageDecoderPlugin::initialize()
{
return {};
}
bool JPEGImageDecoderPlugin::sniff(ReadonlyBytes data)
{
return data.size() > 3
&& data.data()[0] == 0xFF
&& data.data()[1] == 0xD8
&& data.data()[2] == 0xFF;
}
ErrorOr<NonnullOwnPtr<ImageDecoderPlugin>> JPEGImageDecoderPlugin::create(ReadonlyBytes data)
{
auto stream = TRY(try_make<FixedMemoryStream>(data));
return adopt_nonnull_own_or_enomem(new (nothrow) JPEGImageDecoderPlugin(move(stream)));
}
bool JPEGImageDecoderPlugin::is_animated()
{
return false;
}
size_t JPEGImageDecoderPlugin::loop_count()
{
return 0;
}
size_t JPEGImageDecoderPlugin::frame_count()
{
return 1;
}
size_t JPEGImageDecoderPlugin::first_animated_frame_index()
{
return 0;
}
ErrorOr<ImageFrameDescriptor> JPEGImageDecoderPlugin::frame(size_t index)
{
if (index > 0)
return Error::from_string_literal("JPEGImageDecoderPlugin: Invalid frame index");
if (m_context->state == JPEGLoadingContext::State::Error)
return Error::from_string_literal("JPEGImageDecoderPlugin: Decoding failed");
if (m_context->state < JPEGLoadingContext::State::BitmapDecoded) {
if (auto result = decode_jpeg(*m_context); result.is_error()) {
m_context->state = JPEGLoadingContext::State::Error;
return result.release_error();
}
m_context->state = JPEGLoadingContext::State::BitmapDecoded;
}
return ImageFrameDescriptor { m_context->bitmap, 0 };
}
ErrorOr<Optional<ReadonlyBytes>> JPEGImageDecoderPlugin::icc_data()
{
TRY(decode_header(*m_context));
if (m_context->icc_data.has_value())
return *m_context->icc_data;
return OptionalNone {};
}
}
|