Probe and report structure and metadata content of camera image files
exifprobe [options] filename(s)
Exifprobe reads image files produced by digital cameras (including several so-called "raw" file formats) and reports the structure of the files and the auxiliary data and metadata contained within them. In addition to TIFF, JPEG, and EXIF, the program understands several formats which may contain "raw" camera data, including MRW, CIFF/CRW, JP2/JPEG2000, RAF, and X3F, as well as most most TIFF-derived "raw" formats, including DNG, ORF, CR2, NEF, K25/KDC/DCR, and PEF. Other TIFF or JPEG-derived formats (which begin with a TIFF header or JPEG marker) should also be readable. The program attempts to display ALL information in the image file, in a manner which mimics the structure of the file as closely as possible.
Where possible, output is not limited to “known” data items. I.e. for tagged file formats, unknown tags will be identified by tag number and type, and values shown without interpretation. Proprietary, untagged or fixed format files do not permit this, but unknown or undefined data can usually be dumped in a hex/ascii format for examination, so long as the file structure is understood. The program will report the contents of any properly structured TIFF IFD or CIFF directory encountered, even when entry tags are not recognized. Recognized TIFF, TIFF/EP, DNG, and CIFF tags are expanded, including EXIF2.2 sections and camera MakerNotes which are found to be in TIFF IFD format. TIFF and/or JPEG sections found in MRW, RAF or JP2 files will be reported, along with the “native” sections of those formats. JP2 boxes will be reported whether known or unknown, and expanded if known. Unknown boxes cannot be expanded, since JP2 is not tagged below the box (or sub-box) level.
An effort is made to identify all sub-images (in any image format) contained in multi-image files; the location, size, and format of such images is reported, and a hex/ascii dump of the image data may be requested. Image data is not processed, but the program will recognize and report all standard JPEG and JPEG2000 markers (including JPEG APPn markers) and will expand APP0 (JFIF/JFXX) and APP1 (EXIF) sections.
Since the program does not attempt to display images or modify the contents of files, it can often recover from and report failures or warn about structural oddities which would confuse a display or image edit program.
There are a wide variety of output formats, selectable in detail by the “lower case” options described in the OPTIONS section below. These options select which items to print, and within narrow confines, how and where to print them. A large number of combinations of options exist, and some of those combinations may not be sensible. In order to avoid the need for constant creativity or invention, three “prefabricated” output formats are provided, selected by the upper case option letters.
The structural (-S) output format (default) provides a description of the image file which mimics as closely as possible the layout and structure of the data in the image file, including file offsets of headers, section and segment markers, fully described TIFF IFDs, CIFF directories, or JP2 boxes, and the location of actual image and/or thumbnail data. The contents of each section are indented relative to the beginning of the section, and ”offset” values for TIFF IFDs and CIFF directories are reported at the offsets where they are found (usually following the entry list for TIFF, or in the HEAP for CIFF). The peculiar “reverse” structures of CIFF and X3F formats are handled sensibly.
The report format (-R) shows the “logical” structure of the image file, but eliminates addresses, offsets, IFD value types and counts, etc., and prints “offset” directory values inline, while otherwise preserving the primary structure of the data.
The list format (-L) omits all structural data. It writes only “tag” values from TIFF, Exif, and MakerNote IFDs, CIFF or other format directories or JP2 boxes, including section and image offsets and sizes. Identifiable values from non-tagged formats are written in a similar manner. This format may be useful for extracting information for photo galleries.
The structural format is default. This format provides maximum information about the contents of an image file, and may reveal information (sometimes important) which other formats (or image info programs) may hide.
In all formats, the filename, file type, file size, an image summary, and a summary file format will be displayed (even when all other output is disabled by option). The image summary includes a summary report, for each subimage found, giving the image type, compression type (if any), pixel size, data length, file offset where found, and section of the file which includes or references the image. In some cases, short remarks may be included for images mentioned but not found, etc. The summary concludes with the number of images found, and number of images not found (if any). The summary is followed by a listing of format sections found (TIFF/JPEG/EXIF, etc.) and a type identifier for TIFF-derived types (e.g. CR2) where possible.
An environment variable may be set to a list of options to customize the default output behavior (see below).
Camera-generated images which contain EXIF sections may also contain sections introduced by a MakerNote tag, which may contain information about camera or firmware settings used to produce the image. The structure and contents of MakerNote sections is not mandated by the Exif specification, but many camera MakerNotes are written in TIFF IFD format, possibly offset following an ID string or new TIFF header (or both), and sometimes with inventive handling of “offsets”. Exifprobe currently understands and automatically detects such schemes and prints the contents of the IFD (and the ID string, if present). This detection is not dependent upon make or model of camera. Make and Model information will usually be available from the first TIFF IFD in the file; this information may be used to interpret the MakerNote information for “known” cameras; otherwise, tag numbers, sizes, types, and raw values from the IFD will be shown (if permitted by option settings).
Some camera makes are known to use more than one version of MakerNote, depending upon model. If an unknown model from that maker is encountered, the note will be briefly examined and a noteversion assigned automatically if possible. If that fails, the note will be displayed without interpretation.
MakerNotes which are not in a recognizable IFD format will be reported (start and end offsets) in structural (-S) and report (-R) formats , and the beginning of the note section hex/ascii dumped. The remainder of the note may be dumped, in whole or in part, by the -M option (see below).
In list (-L) format, the starting file offset and length supplied by the MakerNote tag will be reported, and three “pseudo” tags which report the offset (MakerNote.Offset), size (MakerNote.Length) and scheme (MakerNote.Scheme) will appear.
In JPEG interchange format files, APP0 (JFIF,JFXX) and APP1 (Exif) segments will be fully decoded, and the “printable” portions of APP12 sections will be displayed. APP3 (Meta) sections will be expanded and the contained TIFF IFD will be displayed, although little interpretation is done. Other APP markers will be reported, and the sections may be hex/ascii dumped in whole or in part using the -A option. APP1 sections not marked as Exif will be treated as unknown.
The program (by default) emits ANSI color escape sequences to highlight Exif, MakerNote , and Interoperability sub-sections. Errors and warnings are highlighted in red. These sequences are effective, of course, only for terminals or terminal emulators (e.g. xterm) which respond to ANSI color escape sequences. If a pager is used to display the output when these sequences are present, a “raw” option to the pager may be required (e.g. less -R). The use of these sequences may be toggled by the -c option. LIST mode turns color sequences off.
The program may be compiled without support for color sequences.
In all formats, and regardless of option setting, the first three lines of output for each file processed are the filename, image type (TIFF, CIFF, JP2, etc.), and the file size. If the type (taken from the file header) specifies a data byte order, the byte order will be indicated with the type as `II' (Intel byte order) or `MM' (Motorola byte order). The image summary and summary format will always be printed at the end.
Structural format output begins with a display of the file header. The header is followed by lines of the form
<IFD0>
...
</IFD0>
<APP0>
...
</APP0>
<DIRECTORY>
...
</DIRECTORY>
etc. to indicate the beginning and end of each “section” of the file. Actual section names will, of course, depend upon the file format and/or the tags encountered. Only the TIFF IFD format is described here; other formats are similar, except that JP2 box names are printed inside square (rather than angle) brackets, and MRW section names inside curly braces.
Within sections, directory entries, subdirectories, the contents of known APP sections, JPEG segment markers, etc. are printed. Non-jpeg image data sections will be shown with a few lines of hex/ascii dump of the beginning of the data.
Each line of output is preceded by a file offset given in hex and decimal. File offsets are preceded by the character `@', except that section end markers are preceded by `-' and the character `>' may be used to mark sections which are located outside the IFD in which they are declared. If that section includes a subsection which is similarly afflicted, the '>' is replaced by '+' in the subsection. In JP2 files, the '@' is replaced by '=', for no particular reason.
JPEG and JPEG2000 segment markers are written with the marker name, and the decoded values of any information associated with the marker.
TIFF information is written in a manner which reflects the structure of the IFD, with all values interpreted according to the applicable specification where possible. All IFD fields are reported. The following fields will appear on each line (in the order given, following the file offset):
Tag number in hex and decimal representations, enclosed in brackets.
Tag name (where known); names for unknown tags are created as a hex representation of the tag number prefixed by the string 'TAG_'.
The TIFF type number, name, and byte count for the associated value, enclosed in square brackets.
The “value/offset” for the entry. If the value fits in the four bytes of the entry, the value is printed directly.
If the value for the entry did not fit in the four bytes of the entry, then the value found is an offset to the actual location of the data; that offset is printed preceded by an '@' symbol. The actual value will be printed later, at the file offset where it was found (except in some non-conforming MakerNote IFDs). If the value requires interpretation (e.g. TIFF Orientation) it is followed by an '=' sign and the interpretation, enclosed in double quotes (e.g. “0,0 top left”).
The list of entries will be followed by a line giving the offset to the next IFD (often 0) which is always found at the end of a TIFF IFD entry list.
If there were offset entries found in the list above, the TIFF (and Exif) specification requires that they will be located next in the file, immediately following the dirctory entries. This stricture is frequently ignored in MakerNotes and TIFF-derived formats. A line reporting the beginning of these offset values will be printed immediately after the next IFD offset, followed by one line for each offset entry, with the tag name repeated, followed by the actual value, followed by its interpretation (if any).
Multiple values in entries are printed on a single line, but large lists will be elided, with just the first two or three values shown, followed by an ellipsis, followed by the last value, the number of values, and the offset of the last value in the list. The full value list may be printed using the -eA option.
In structural format, ascii strings in the entry are printed for the entire length given in the IFD entry, including nulls and non-ascii values (if present), which are printed in `backslashed' octal notation. The -ea option may be used to force ascii values to be printed only up to the first null. This option is often necessary for CIFF format files, and is enabled by default in “list” mode.
Entries are indented slightly from the start identifier for the IFD, and subsegments (e.g. an Exif IFD, SubIFD, or MakerNote) will be further indented in order to indicate the structure of the file.
The resulting output displays the contents of the IFD much as it appears in the file (see the TIFF or EXIF specifications for descriptions of the IFD format).
Finally, the start and end of actual image data for the primary image (and possibly thumbnail or reduced-resolution image) is reported at the end. For JPEG images, this usually includes display of the JPEG segment markers within the image. Binary format image data will be shown with a brief hex/ascii dump of the beginning of the data, between start and end markers.
Note that values preceded by `@' are always offsets from the beginning of the file to the actual value. IFD offsets are usually recorded in the file as offsets relative to the beginning of the TIFF header (which is offset from the beginning of the file in JPEG APP1 files) but are adjusted by exifprobe to show offset from the beginning of the file. If it is important to see the recorded value, the -er option may be used to print the recorded value in parentheses, following the adjusted file offset.
The report format (-R) displays all sections and segments of the image file, including start and end of sections, but eliminates much of the “cruft” of the structural format by eliminating address/offset information and much of the `internal' information from the TIFF IFD (tag number, type and count). Offset values are printed inline with the tag name. The output is indented to show the logical structure of the image file, but is much less difficult to view than the structural format.
The list format (-L) suppresses structural information, writing only content in the format tagname = value or tagname = value = “what value means”. For non-tagged file formats, the tagname will be replaced by a fixed identifier for the item. In LIST format, “long” tagnames are used, which include the names of all parent sections of the section in which the data is found. Long tagnames can be toggled off, although this is unwise if the file contains multiple image sections.
The “value” of tags or items which represent an offset to a subsection or image are printed in list format as “@offset:length”.
The List format is used by the auxiliary script exifgrep, which permits selective extraction of information e.g. for photo galleries, and output in (almost) “shell variable” format.
The -Z option “zeroes” all option flags (except the longnames modifier), after which the lower-case options may be used to set desired options. The lower-case options are `toggles', which may also be used to turn off items in the pre-defined formats.
As an example, the command:
exifprobe -Z -et somefile.tif
may be used to list just the TIFF and Exif tags, without values (or anything else) in “long name” format.
exifprobe -eNnT somefile.tif
will print in structural format, suppressing output of hex and decimal tag numbers, and tag type and count.
The “zero-level” output still reports the file data and image summary as described above.
The environment variable EXIFPROBE_OPTIONS may be set to any valid option string, which will be evaluated before command line options. E.g.
export EXIFPROBE_OPTIONS='-L -c'
will make list format the default output format, and re-enable the color sequences turned off by -L.
Options are evaluated from left to right, so -Z should be given first, while -D (decimal only) or -X (hex only) should be given last.
Structure mode: (default) almost everything; offset values not inline
Report mode: like structural, but only tagnames and decimal values, indented, inline
List mode: print only tags and values (including interpreted values); no section info; no color
turn off (zero) all optional output. Prints only filename, filetype, filesize, image summary, and file format.
toggle use of ANSI color control sequences to emphasize EXIF sections. (default 'on' except list mode, unless compiled with no color support)
toggle printing of addresses (file offsets) in hex and decimal
three-way toggle indent (after address -> before -> none)
toggle “inline” print of offset IFD values
toggle print identifiers for:
- sections (IFDs, APPn)
- segments (JPEG segments)
- JPEG APP0...APPn entries
- long names (dot-separated list of parent sections preceding item name)
- entries. Includes tag names, numbers, types, values.
- print MakerNote scheme description
- watch debug of MakerNote scheme detection
toggle print IFD entry items:
- tagname
- tag number in decimal
- tag number in hex
- entry type and count
- value in decimal
- value in hex
- file offset to value in decimal
- file offset to value in hex
- relative (unadjusted) offset in decimal
- also print “raw” values where normal values are computed (e.g. rational values, or some MakerNote values where APEX values must be computed from a raw value).
- print ALL elements of multiple-value tags
- ascii "ignore length" (stop at first null)
limit all enabled numerical values to decimal only (addresses, tag numbers, offsets, values)
limit all enabled numerical values to hex only (addresses, tag numbers, offsets, values)
dump len (or all) bytes of UNDEFINED data found in TIFF IFDS in hex/ascii form (but only if the structure of the data is not known)
dump len (or all) bytes of unrecognized MakerNotes in hex/ascii form (but only if the structure of the data is not known)
dump len (or all) bytes of unrecognized JPEG APP segments in hex/ascii form (but only if the structure of the data is not known)
dump len (or all) bytes of binary image data or failed JPEG image data
print a list of camera makes/models matching make or model as substrings. `+' by itself prints everything
start processing at file offset start_offset
print filename at beginning of each line of output (useful when grepping multiple files in LIST mode)
force use of note version noteversion when interpreting MakerNotes. Useful only if you know what you're doing.
Force the makernote code to interpret the note according to the make given, rather than that contained in the file.
force the makernote code to interpret the note according to the model given, rather than that contained in the file.
This option has effect only if set in EXIFPROBE_OPTIONS. If set when command line options are processed, color will be be off by default if the output is not to a tty. Any command line option which toggles or sets color (e.g. “-R”) will turn color back on.
Print “raw” Unicode data. Normally 16 bit data is printed as though the high byte is zero (which is often the case). Writing the nulls would annoy most ascii terminal devices, so the default is more hospitable. The -u option forces printing of the full value.
print a help message
print program version and copyright
The TIFF6 specification:
https://partners.adobe.com/asn/developer/PDFS/TN/TIFF6.pdf
The Exif 2.2 specification:
http://tsc.jeita.or.jp/avs/data/cp3451.pdf
The JFIF specification:
http://www.w3.org/Graphics/JPEG/jfif3.pdf
The TIFF/EP specification:
http://www.map.tu.chiba-u.ac.jp/IEC/100/TA2/recdoc/N4378.pdf
The CIFF specification http://xyrion.org/ciff/CIFFspecV1R04.pdf
The X3F public specification http://www.sd9.org.uk/X3F_Format.pdf
The JPEG2000 public draft (outdated) http://www.jpeg.org/public/fcd15444-1.pdf
Most diagnostics are printed “inline” to stdout, in red if color is enabled, and the program attempts to proceed.
Interpretation of MakerNote data for specific cameras is incomplete and probably always will be. The X3F specification is incomplete, and the final JPEG2000/JP2 specification is not freely available; support for these formats is therefore not complete, and may not be entirely accurate.
The RAF file format support is preliminary (there is no published specification).
Floating point values read from the file are expected to be in IEEE format (or at least, native format); i.e. no conversions are attempted.
ANSI color sequence support should use termcap/terminfo facilities; it does not.
Duane H. Hesser