1 2#------------------------------------------------------------------------------ 3# JPEG images 4# SunOS 5.5.1 had 5# 6# 0 string \377\330\377\340 JPEG file 7# 0 string \377\330\377\356 JPG file 8# 9# both of which turn into "JPEG image data" here. 10# 110 beshort 0xffd8 JPEG image data 12!:mime image/jpeg 13!:apple 8BIMJPEG 14!:strength +1 15>6 string JFIF \b, JFIF standard 16# The following added by Erik Rossen <rossen@freesurf.ch> 1999-09-06 17# in a vain attempt to add image size reporting for JFIF. Note that these 18# tests are not fool-proof since some perfectly valid JPEGs are currently 19# impossible to specify in magic(4) format. 20# First, a little JFIF version info: 21>>11 byte x \b %d. 22>>12 byte x \b%02d 23# Next, the resolution or aspect ratio of the image: 24#>>13 byte 0 \b, aspect ratio 25#>>13 byte 1 \b, resolution (DPI) 26#>>13 byte 2 \b, resolution (DPCM) 27#>>4 beshort x \b, segment length %d 28# Next, show thumbnail info, if it exists: 29>>18 byte !0 \b, thumbnail %dx 30>>>19 byte x \b%d 31 32# EXIF moved down here to avoid reporting a bogus version number, 33# and EXIF version number printing added. 34# - Patrik R=E5dman <patrik+file-magic@iki.fi> 35>6 string Exif \b, EXIF standard 36# Look for EXIF IFD offset in IFD 0, and then look for EXIF version tag in EXIF IFD. 37# All possible combinations of entries have to be enumerated, since no looping 38# is possible. And both endians are possible... 39# The combinations included below are from real-world JPEGs. 40# Little-endian 41>>12 string II 42# IFD 0 Entry #5: 43>>>70 leshort 0x8769 44# EXIF IFD Entry #1: 45>>>>(78.l+14) leshort 0x9000 46>>>>>(78.l+23) byte x %c 47>>>>>(78.l+24) byte x \b.%c 48>>>>>(78.l+25) byte !0x30 \b%c 49# IFD 0 Entry #9: 50>>>118 leshort 0x8769 51# EXIF IFD Entry #3: 52>>>>(126.l+38) leshort 0x9000 53>>>>>(126.l+47) byte x %c 54>>>>>(126.l+48) byte x \b.%c 55>>>>>(126.l+49) byte !0x30 \b%c 56# IFD 0 Entry #10 57>>>130 leshort 0x8769 58# EXIF IFD Entry #3: 59>>>>(138.l+38) leshort 0x9000 60>>>>>(138.l+47) byte x %c 61>>>>>(138.l+48) byte x \b.%c 62>>>>>(138.l+49) byte !0x30 \b%c 63# EXIF IFD Entry #4: 64>>>>(138.l+50) leshort 0x9000 65>>>>>(138.l+59) byte x %c 66>>>>>(138.l+60) byte x \b.%c 67>>>>>(138.l+61) byte !0x30 \b%c 68# EXIF IFD Entry #5: 69>>>>(138.l+62) leshort 0x9000 70>>>>>(138.l+71) byte x %c 71>>>>>(138.l+72) byte x \b.%c 72>>>>>(138.l+73) byte !0x30 \b%c 73# IFD 0 Entry #11 74>>>142 leshort 0x8769 75# EXIF IFD Entry #3: 76>>>>(150.l+38) leshort 0x9000 77>>>>>(150.l+47) byte x %c 78>>>>>(150.l+48) byte x \b.%c 79>>>>>(150.l+49) byte !0x30 \b%c 80# EXIF IFD Entry #4: 81>>>>(150.l+50) leshort 0x9000 82>>>>>(150.l+59) byte x %c 83>>>>>(150.l+60) byte x \b.%c 84>>>>>(150.l+61) byte !0x30 \b%c 85# EXIF IFD Entry #5: 86>>>>(150.l+62) leshort 0x9000 87>>>>>(150.l+71) byte x %c 88>>>>>(150.l+72) byte x \b.%c 89>>>>>(150.l+73) byte !0x30 \b%c 90# Big-endian 91>>12 string MM 92# IFD 0 Entry #9: 93>>>118 beshort 0x8769 94# EXIF IFD Entry #1: 95>>>>(126.L+14) beshort 0x9000 96>>>>>(126.L+23) byte x %c 97>>>>>(126.L+24) byte x \b.%c 98>>>>>(126.L+25) byte !0x30 \b%c 99# EXIF IFD Entry #3: 100>>>>(126.L+38) beshort 0x9000 101>>>>>(126.L+47) byte x %c 102>>>>>(126.L+48) byte x \b.%c 103>>>>>(126.L+49) byte !0x30 \b%c 104# IFD 0 Entry #10 105>>>130 beshort 0x8769 106# EXIF IFD Entry #3: 107>>>>(138.L+38) beshort 0x9000 108>>>>>(138.L+47) byte x %c 109>>>>>(138.L+48) byte x \b.%c 110>>>>>(138.L+49) byte !0x30 \b%c 111# EXIF IFD Entry #5: 112>>>>(138.L+62) beshort 0x9000 113>>>>>(138.L+71) byte x %c 114>>>>>(138.L+72) byte x \b.%c 115>>>>>(138.L+73) byte !0x30 \b%c 116# IFD 0 Entry #11 117>>>142 beshort 0x8769 118# EXIF IFD Entry #4: 119>>>>(150.L+50) beshort 0x9000 120>>>>>(150.L+59) byte x %c 121>>>>>(150.L+60) byte x \b.%c 122>>>>>(150.L+61) byte !0x30 \b%c 123# Here things get sticky. We can do ONE MORE marker segment with 124# indirect addressing, and that's all. It would be great if we could 125# do pointer arithemetic like in an assembler language. Christos? 126# And if there was some sort of looping construct to do searches, plus a few 127# named accumulators, it would be even more effective... 128# At least we can show a comment if no other segments got inserted before: 129>(4.S+5) byte 0xFE 130>>(4.S+8) string >\0 \b, comment: "%s" 131# FIXME: When we can do non-byte counted strings, we can use that to get 132# the string's count, and fix Debian bug #283760 133#>(4.S+5) byte 0xFE \b, comment 134#>>(4.S+6) beshort x \b length=%d 135#>>(4.S+8) string >\0 \b, "%s" 136# Or, we can show the encoding type (I've included only the three most common) 137# and image dimensions if we are lucky and the SOFn (image segment) is here: 138>(4.S+5) byte 0xC0 \b, baseline 139>>(4.S+6) byte x \b, precision %d 140>>(4.S+7) beshort x \b, %dx 141>>(4.S+9) beshort x \b%d 142>(4.S+5) byte 0xC1 \b, extended sequential 143>>(4.S+6) byte x \b, precision %d 144>>(4.S+7) beshort x \b, %dx 145>>(4.S+9) beshort x \b%d 146>(4.S+5) byte 0xC2 \b, progressive 147>>(4.S+6) byte x \b, precision %d 148>>(4.S+7) beshort x \b, %dx 149>>(4.S+9) beshort x \b%d 150# I've commented-out quantisation table reporting. I doubt anyone cares yet. 151#>(4.S+5) byte 0xDB \b, quantisation table 152#>>(4.S+6) beshort x \b length=%d 153#>14 beshort x \b, %d x 154#>16 beshort x \b %d 155 156# HSI is Handmade Software's proprietary JPEG encoding scheme 1570 string hsi1 JPEG image data, HSI proprietary 158 159# From: David Santinoli <david@santinoli.com> 1600 string \x00\x00\x00\x0C\x6A\x50\x20\x20\x0D\x0A\x87\x0A JPEG 2000 image data 161 162# Type: JPEG 2000 codesream 163# From: Mathieu Malaterre <mathieu.malaterre@gmail.com> 1640 belong 0xff4fff51 JPEG 2000 codestream 16545 beshort 0xff52 166